zaiweizhang / h3dnet Goto Github PK

Thanks for the code, nice work. Would you share the 'viewpoint.json' file which used in 'show_results_scannet.py'? When I tried to run the code I found the code cannot find the 'viewpoint.json'.

Hi @zaiweizhang , @GitBoSun
Please provide tips on how to use H3DNet on our custom dataset, one having very dfferent distribution.
Specifically how to properly tune/choose the config parameters and threshold values?

Hi,

Thank you for sharing this great work!
I have a simple question about the performance with single backbone.
What is the [email protected] and [email protected] on both ScanNet and SUN RGB-D when the model use single backbone?

Regards,

Hi, I find that there may be some mistakes in PointnetSAModuleMatch module.

In file proposal_module_refine.py:
line 328:
surface_xyz, surface_features, _ = self.match_surface_center(torch.cat((obj_surface_center, surface_center_pred), dim=1), torch.cat((obj_surface_feature, surface_center_feature_pred), dim=2))

In line 119: match_surface_center is defined as:

### surface center matching
       self.match_surface_center = PointnetSAModuleMatch( 
            npoint=self.num_proposal*6,
            radius=0.5,
            nsample=32,
            mlp=[128+6, 128, 64, 32],
            use_xyz=True,
            normalize_xyz=True
        )

As for torch.cat((obj_surface_center, surface_center_pred), dim=1) and torch.cat((obj_surface_feature, surface_center_feature_pred), dim=2), we can get a set of points and their correspondence features.

Then, I go further into the source code of PointnetSAModuleMatch module:
pointnet2_modules.py: Line 467:

        new_xyz = xyz[:,:self.npoint,:].contiguous()
        target_xyz = xyz[:,self.npoint:,:].contiguous()
        
        if not self.ret_unique_cnt:
            grouped_features, grouped_xyz = self.grouper(
                target_xyz, new_xyz, features
            )  # (B, C, npoint, nsample)
        else:
            grouped_features, grouped_xyz, unique_cnt = self.grouper(
                target_xyz, new_xyz, features
            )  # (B, C, npoint, nsample), (B,3,npoint,nsample), (B,npoint)

Here, I think the difference between PointnetSAModuleMatch and original PointnetSAModule is that user can specify new_xyz in PointnetSAModuleMatch module.
However, I found some problems in Line 472. target_xyz and features are two parameters parsed to function self.grouper, but they are mismatch in these case. Problem will happen while grouping the neighbour features, because the features of target_xyz is begin from features[:,:,self.npoint:]. I think we should correcte this code with

grouped_features, grouped_xyz, unique_cnt = self.grouper(
                target_xyz, new_xyz, features[:,:,self.npoint:].contiguous()
            )

Hi, is it possible to provide the model (ScanNet and SUN RGB-D) reported in your paper?

Thanks very much.

Hi, thanks for your great job. Here is a question about the training process of SUN-RGBD.
I noticed that the axis is aligned to the gravity direction. The tilt angle is provided in your processed dataset. This means if I want to test the model with a new RGBD image taken from Kinect v2 sensor, I will also have to calculate the tilt angle. Do you have any idea how to do this? Are there any tools out of the box?

Any suggestions will be helpful. Thanks very much!

Hi~
My machine is in-built 7700K CPU and a 1080ti GPU, it took 15 mins for one epoch training. It took more than 90 hours to complete training. Can you tell me about the training time and the GPU of your machine?

Hi bro @zaiweizhang ,

I am trying to train H3DNet on Stanford 3D-Semantics Dataset. When I convert Stanford 3D-Semantics Dataset to Scannet V2 format, I face a difficult for "sence_all_noangle_40cls.npy". Could you please share with me your code for convert to "sence_all_noangle_40cls.npy" ? As I read your code and VoteNet, column 8th at file "all_noangle_40cls.npy" is instance_labels and column 9th is semantic_labels. But I do not know column from 1st to 7th is represent for which data.

I read from Votenet Code column from 1st to 7th seem is a box as (cx,cy,cz,dx,dy,dz,semantic_label) but I am not sure because a file "noangle_40cls.npy" has (50k ,9) shape, so I think boxes can not be represent as format like this.

I hope to receive your advice soon. Thank you very much.

Hi~How do you get the ground truth in Figure 10?

Hi @zaiweizhang , @GitBoSun

How can I run the model for detecting objects on my custom data/images? The classes can still be the same as scannet/sunrgbd dataset for now. From what i understand based on looking at sunrgbd data:
For evaluation
I need 3 files -

1. bbox.npy - this contains 3D bboxes of objects in the scene
2. pc.npz - this contains the point cloud
3. votez.npz - this contains a Nx10 array describing votes (from votenet?) that is used for detection.

lets say i capture an RGBD image. i can fill in the depth image and get a dense pointcloud (along with color).
How/what can i do to run the trained model on this file.
I have 2). 1) should only be used for evaluation and not inference.
How do i get 3)?

Hi, I try to training the H3DNet on SUN-RGBD dataset(using the default configs), but my reproduced performance is lower than the paper reported. Are there others tricks that I missed?

Hi zaiwei,

When performing data augmentation on ScanNet, you also flipped the height.

Additionally, you set the heading angle as the rot_angle (since original heading angle is 0 in ScanNet)

Thank you.

When I trained model as your guideline , this bug happen. I tried to install step.py at pointnet2 folder but it can't install. Please help me to check . thank you

2021-02-05 02:46:53.256615: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1
5285 5050
331 316
Traceback (most recent call last):
File "/content/drive/My Drive/Computer_Vision/H3DNet/H3DNet/pointnet2/pointnet2_utils.py", line 26, in
import pointnet2._ext as _ext
ModuleNotFoundError: No module named 'pointnet2._ext'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "train_1bb.py", line 171, in
MODEL = importlib.import_module(FLAGS.model) # import network module
File "/usr/lib/python3.6/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "", line 994, in _gcd_import
File "", line 971, in _find_and_load
File "", line 955, in _find_and_load_unlocked
File "", line 665, in _load_unlocked
File "", line 678, in exec_module
File "", line 219, in _call_with_frames_removed
File "/content/drive/My Drive/Computer_Vision/H3DNet/H3DNet/models/hdnet_1bb.py", line 23, in
from backbone_module_scale import Pointnet2Backbone
File "/content/drive/My Drive/Computer_Vision/H3DNet/H3DNet/models/backbone_module_scale.py", line 19, in
from pointnet2_modules import PointnetSAModuleVotes, PointnetSAModuleVotesWith, PointnetFPModule, PointnetPlaneVotes
File "/content/drive/My Drive/Computer_Vision/H3DNet/H3DNet/pointnet2/pointnet2_modules.py", line 21, in
import pointnet2_utils
File "/content/drive/My Drive/Computer_Vision/H3DNet/H3DNet/pointnet2/pointnet2_utils.py", line 30, in
"Could not import _ext module.\n"
ImportError: Could not import _ext module.
Please see the setup instructions in the README: https://github.com/erikwijmans/Pointnet2_PyTorch/blob/master/README.rst

Dear author,

Thanks for your great work!

I have two problems about your pre-processed data.

1. As indicated in your readme, the pre-processing procedure is same with votenet. However, after processed by votenet code, there will be no _all_noangle_40cls.npy for producing the meta_vertice in your code here.

2. When I tried to download your processed data, the link you provide shows "file does not exist".

Could you please help with this two questions?

Thanks a lot!

Hi Zhang,

I am trying to finetune your model on sub dataset of Scannet V2 ( I pick only most 3 popular classes ). Do you have any suggestion for me ? I tried to freeze your weight and only train last layer but the mAP was not increased.

Thank you very much,

I execute this code on Ubuntu16.04,then there is an error:OSError: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.27' not found.Is there any way to deal with this bug,except upgrading the system.Wish ur help,plz.

Hi,

I find that here, += may cause inplace error in many versions of PyTorch.

You may revise it by "net = net + original_feature".

Hi, thanks for sharing your code.

I am trying to reproduce the results on ScanNet using the default configs. But I can only obtain around 64-65 [email protected], which is lower than the results in your paper (67.2). Do you have any advice to reproduce your results?

Here are the training logs log_train.txt log_train_run2.txt

Hi Zhang,

I train H3D on 1 GPU, everything is okay. But when I move to train on multi GPUs (4 GPUs), a issue happens .
I tried the solution load layer as order as the guide bearpaw/pytorch-classification#27 but the issue still happens. Do you have any idea about this issue ? Thank you very much.

net.load_state_dict(new_state_dict['module.model_state_dict']) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 845, in load_state_dict self.class.name, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for DataParallel: Missing key(s) in state_dict: "module.backbone_net1.sa1.mlp_module.layer0.conv.weight", "module.backbone_net1.sa1.mlp_module.layer0.bn.bn.weight", "module.backbone_net1.sa1.mlp_module.layer0.bn.bn.bias", "module.backbone_net1.sa1.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net1.sa1.mlp_module.layer0.bn.bn.running_var", "module.backbone_net1.sa1.mlp_module.layer1.conv.weight", "module.backbone_net1.sa1.mlp_module.layer1.bn.bn.weight", "module.backbone_net1.sa1.mlp_module.layer1.bn.bn.bias", "module.backbone_net1.sa1.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net1.sa1.mlp_module.layer1.bn.bn.running_var", "module.backbone_net1.sa1.mlp_module.layer2.conv.weight", "module.backbone_net1.sa1.mlp_module.layer2.bn.bn.weight", "module.backbone_net1.sa1.mlp_module.layer2.bn.bn.bias", "module.backbone_net1.sa1.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net1.sa1.mlp_module.layer2.bn.bn.running_var", "module.backbone_net1.sa2.mlp_module.layer0.conv.weight", "module.backbone_net1.sa2.mlp_module.layer0.bn.bn.weight", "module.backbone_net1.sa2.mlp_module.layer0.bn.bn.bias", "module.backbone_net1.sa2.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net1.sa2.mlp_module.layer0.bn.bn.running_var", "module.backbone_net1.sa2.mlp_module.layer1.conv.weight", "module.backbone_net1.sa2.mlp_module.layer1.bn.bn.weight", "module.backbone_net1.sa2.mlp_module.layer1.bn.bn.bias", "module.backbone_net1.sa2.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net1.sa2.mlp_module.layer1.bn.bn.running_var", "module.backbone_net1.sa2.mlp_module.layer2.conv.weight", "module.backbone_net1.sa2.mlp_module.layer2.bn.bn.weight", "module.backbone_net1.sa2.mlp_module.layer2.bn.bn.bias", "module.backbone_net1.sa2.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net1.sa2.mlp_module.layer2.bn.bn.running_var", "module.backbone_net1.sa3.mlp_module.layer0.conv.weight", "module.backbone_net1.sa3.mlp_module.layer0.bn.bn.weight", "module.backbone_net1.sa3.mlp_module.layer0.bn.bn.bias", "module.backbone_net1.sa3.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net1.sa3.mlp_module.layer0.bn.bn.running_var", "module.backbone_net1.sa3.mlp_module.layer1.conv.weight", "module.backbone_net1.sa3.mlp_module.layer1.bn.bn.weight", "module.backbone_net1.sa3.mlp_module.layer1.bn.bn.bias", "module.backbone_net1.sa3.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net1.sa3.mlp_module.layer1.bn.bn.running_var", "module.backbone_net1.sa3.mlp_module.layer2.conv.weight", "module.backbone_net1.sa3.mlp_module.layer2.bn.bn.weight", "module.backbone_net1.sa3.mlp_module.layer2.bn.bn.bias", "module.backbone_net1.sa3.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net1.sa3.mlp_module.layer2.bn.bn.running_var", "module.backbone_net1.sa4.mlp_module.layer0.conv.weight", "module.backbone_net1.sa4.mlp_module.layer0.bn.bn.weight", "module.backbone_net1.sa4.mlp_module.layer0.bn.bn.bias", "module.backbone_net1.sa4.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net1.sa4.mlp_module.layer0.bn.bn.running_var", "module.backbone_net1.sa4.mlp_module.layer1.conv.weight", "module.backbone_net1.sa4.mlp_module.layer1.bn.bn.weight", "module.backbone_net1.sa4.mlp_module.layer1.bn.bn.bias", "module.backbone_net1.sa4.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net1.sa4.mlp_module.layer1.bn.bn.running_var", "module.backbone_net1.sa4.mlp_module.layer2.conv.weight", "module.backbone_net1.sa4.mlp_module.layer2.bn.bn.weight", "module.backbone_net1.sa4.mlp_module.layer2.bn.bn.bias", "module.backbone_net1.sa4.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net1.sa4.mlp_module.layer2.bn.bn.running_var", "module.backbone_net1.fp1.mlp.layer0.conv.weight", "module.backbone_net1.fp1.mlp.layer0.bn.bn.weight", "module.backbone_net1.fp1.mlp.layer0.bn.bn.bias", "module.backbone_net1.fp1.mlp.layer0.bn.bn.running_mean", "module.backbone_net1.fp1.mlp.layer0.bn.bn.running_var", "module.backbone_net1.fp1.mlp.layer1.conv.weight", "module.backbone_net1.fp1.mlp.layer1.bn.bn.weight", "module.backbone_net1.fp1.mlp.layer1.bn.bn.bias", "module.backbone_net1.fp1.mlp.layer1.bn.bn.running_mean", "module.backbone_net1.fp1.mlp.layer1.bn.bn.running_var", "module.backbone_net1.fp2.mlp.layer0.conv.weight", "module.backbone_net1.fp2.mlp.layer0.bn.bn.weight", "module.backbone_net1.fp2.mlp.layer0.bn.bn.bias", "module.backbone_net1.fp2.mlp.layer0.bn.bn.running_mean", "module.backbone_net1.fp2.mlp.layer0.bn.bn.running_var", "module.backbone_net1.fp2.mlp.layer1.conv.weight", "module.backbone_net1.fp2.mlp.layer1.bn.bn.weight", "module.backbone_net1.fp2.mlp.layer1.bn.bn.bias", "module.backbone_net1.fp2.mlp.layer1.bn.bn.running_mean", "module.backbone_net1.fp2.mlp.layer1.bn.bn.running_var", "module.backbone_net2.sa1.mlp_module.layer0.conv.weight", "module.backbone_net2.sa1.mlp_module.layer0.bn.bn.weight", "module.backbone_net2.sa1.mlp_module.layer0.bn.bn.bias", "module.backbone_net2.sa1.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net2.sa1.mlp_module.layer0.bn.bn.running_var", "module.backbone_net2.sa1.mlp_module.layer1.conv.weight", "module.backbone_net2.sa1.mlp_module.layer1.bn.bn.weight", "module.backbone_net2.sa1.mlp_module.layer1.bn.bn.bias", "module.backbone_net2.sa1.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net2.sa1.mlp_module.layer1.bn.bn.running_var", "module.backbone_net2.sa1.mlp_module.layer2.conv.weight", "module.backbone_net2.sa1.mlp_module.layer2.bn.bn.weight", "module.backbone_net2.sa1.mlp_module.layer2.bn.bn.bias", "module.backbone_net2.sa1.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net2.sa1.mlp_module.layer2.bn.bn.running_var", "module.backbone_net2.sa2.mlp_module.layer0.conv.weight", "module.backbone_net2.sa2.mlp_module.layer0.bn.bn.weight", "module.backbone_net2.sa2.mlp_module.layer0.bn.bn.bias", "module.backbone_net2.sa2.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net2.sa2.mlp_module.layer0.bn.bn.running_var", "module.backbone_net2.sa2.mlp_module.layer1.conv.weight", "module.backbone_net2.sa2.mlp_module.layer1.bn.bn.weight", "module.backbone_net2.sa2.mlp_module.layer1.bn.bn.bias", "module.backbone_net2.sa2.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net2.sa2.mlp_module.layer1.bn.bn.running_var", "module.backbone_net2.sa2.mlp_module.layer2.conv.weight", "module.backbone_net2.sa2.mlp_module.layer2.bn.bn.weight", "module.backbone_net2.sa2.mlp_module.layer2.bn.bn.bias", "module.backbone_net2.sa2.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net2.sa2.mlp_module.layer2.bn.bn.running_var", "module.backbone_net2.sa3.mlp_module.layer0.conv.weight", "module.backbone_net2.sa3.mlp_module.layer0.bn.bn.weight", "module.backbone_net2.sa3.mlp_module.layer0.bn.bn.bias", "module.backbone_net2.sa3.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net2.sa3.mlp_module.layer0.bn.bn.running_var", "module.backbone_net2.sa3.mlp_module.layer1.conv.weight", "module.backbone_net2.sa3.mlp_module.layer1.bn.bn.weight", "module.backbone_net2.sa3.mlp_module.layer1.bn.bn.bias", "module.backbone_net2.sa3.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net2.sa3.mlp_module.layer1.bn.bn.running_var", "module.backbone_net2.sa3.mlp_module.layer2.conv.weight", "module.backbone_net2.sa3.mlp_module.layer2.bn.bn.weight", "module.backbone_net2.sa3.mlp_module.layer2.bn.bn.bias", "module.backbone_net2.sa3.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net2.sa3.mlp_module.layer2.bn.bn.running_var", "module.backbone_net2.sa4.mlp_module.layer0.conv.weight", "module.backbone_net2.sa4.mlp_module.layer0.bn.bn.weight", "module.backbone_net2.sa4.mlp_module.layer0.bn.bn.bias", "module.backbone_net2.sa4.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net2.sa4.mlp_module.layer0.bn.bn.running_var", "module.backbone_net2.sa4.mlp_module.layer1.conv.weight", "module.backbone_net2.sa4.mlp_module.layer1.bn.bn.weight", "module.backbone_net2.sa4.mlp_module.layer1.bn.bn.bias", "module.backbone_net2.sa4.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net2.sa4.mlp_module.layer1.bn.bn.running_var", "module.backbone_net2.sa4.mlp_module.layer2.conv.weight", "module.backbone_net2.sa4.mlp_module.layer2.bn.bn.weight", "module.backbone_net2.sa4.mlp_module.layer2.bn.bn.bias", "module.backbone_net2.sa4.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net2.sa4.mlp_module.layer2.bn.bn.running_var", "module.backbone_net2.fp1.mlp.layer0.conv.weight", "module.backbone_net2.fp1.mlp.layer0.bn.bn.weight", "module.backbone_net2.fp1.mlp.layer0.bn.bn.bias", "module.backbone_net2.fp1.mlp.layer0.bn.bn.running_mean", "module.backbone_net2.fp1.mlp.layer0.bn.bn.running_var", "module.backbone_net2.fp1.mlp.layer1.conv.weight", "module.backbone_net2.fp1.mlp.layer1.bn.bn.weight", "module.backbone_net2.fp1.mlp.layer1.bn.bn.bias", "module.backbone_net2.fp1.mlp.layer1.bn.bn.running_mean", "module.backbone_net2.fp1.mlp.layer1.bn.bn.running_var", "module.backbone_net2.fp2.mlp.layer0.conv.weight", "module.backbone_net2.fp2.mlp.layer0.bn.bn.weight", "module.backbone_net2.fp2.mlp.layer0.bn.bn.bias", "module.backbone_net2.fp2.mlp.layer0.bn.bn.running_mean", "module.backbone_net2.fp2.mlp.layer0.bn.bn.running_var", "module.backbone_net2.fp2.mlp.layer1.conv.weight", "module.backbone_net2.fp2.mlp.layer1.bn.bn.weight", "module.backbone_net2.fp2.mlp.layer1.bn.bn.bias", "module.backbone_net2.fp2.mlp.layer1.bn.bn.running_mean", "module.backbone_net2.fp2.mlp.layer1.bn.bn.running_var", "module.backbone_net3.sa1.mlp_module.layer0.conv.weight", "module.backbone_net3.sa1.mlp_module.layer0.bn.bn.weight", "module.backbone_net3.sa1.mlp_module.layer0.bn.bn.bias", "module.backbone_net3.sa1.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net3.sa1.mlp_module.layer0.bn.bn.running_var", "module.backbone_net3.sa1.mlp_module.layer1.conv.weight", "module.backbone_net3.sa1.mlp_module.layer1.bn.bn.weight", "module.backbone_net3.sa1.mlp_module.layer1.bn.bn.bias", "module.backbone_net3.sa1.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net3.sa1.mlp_module.layer1.bn.bn.running_var", "module.backbone_net3.sa1.mlp_module.layer2.conv.weight", "module.backbone_net3.sa1.mlp_module.layer2.bn.bn.weight", "module.backbone_net3.sa1.mlp_module.layer2.bn.bn.bias", "module.backbone_net3.sa1.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net3.sa1.mlp_module.layer2.bn.bn.running_var", "module.backbone_net3.sa2.mlp_module.layer0.conv.weight", "module.backbone_net3.sa2.mlp_module.layer0.bn.bn.weight", "module.backbone_net3.sa2.mlp_module.layer0.bn.bn.bias", "module.backbone_net3.sa2.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net3.sa2.mlp_module.layer0.bn.bn.running_var", "module.backbone_net3.sa2.mlp_module.layer1.conv.weight", "module.backbone_net3.sa2.mlp_module.layer1.bn.bn.weight", "module.backbone_net3.sa2.mlp_module.layer1.bn.bn.bias", "module.backbone_net3.sa2.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net3.sa2.mlp_module.layer1.bn.bn.running_var", "module.backbone_net3.sa2.mlp_module.layer2.conv.weight", "module.backbone_net3.sa2.mlp_module.layer2.bn.bn.weight", "module.backbone_net3.sa2.mlp_module.layer2.bn.bn.bias", "module.backbone_net3.sa2.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net3.sa2.mlp_module.layer2.bn.bn.running_var", "module.backbone_net3.sa3.mlp_module.layer0.conv.weight", "module.backbone_net3.sa3.mlp_module.layer0.bn.bn.weight", "module.backbone_net3.sa3.mlp_module.layer0.bn.bn.bias", "module.backbone_net3.sa3.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net3.sa3.mlp_module.layer0.bn.bn.running_var", "module.backbone_net3.sa3.mlp_module.layer1.conv.weight", "module.backbone_net3.sa3.mlp_module.layer1.bn.bn.weight", "module.backbone_net3.sa3.mlp_module.layer1.bn.bn.bias", "module.backbone_net3.sa3.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net3.sa3.mlp_module.layer1.bn.bn.running_var", "module.backbone_net3.sa3.mlp_module.layer2.conv.weight", "module.backbone_net3.sa3.mlp_module.layer2.bn.bn.weight", "module.backbone_net3.sa3.mlp_module.layer2.bn.bn.bias", "module.backbone_net3.sa3.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net3.sa3.mlp_module.layer2.bn.bn.running_var", "module.backbone_net3.sa4.mlp_module.layer0.conv.weight", "module.backbone_net3.sa4.mlp_module.layer0.bn.bn.weight", "module.backbone_net3.sa4.mlp_module.layer0.bn.bn.bias", "module.backbone_net3.sa4.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net3.sa4.mlp_module.layer0.bn.bn.running_var", "module.backbone_net3.sa4.mlp_module.layer1.conv.weight", "module.backbone_net3.sa4.mlp_module.layer1.bn.bn.weight", "module.backbone_net3.sa4.mlp_module.layer1.bn.bn.bias", "module.backbone_net3.sa4.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net3.sa4.mlp_module.layer1.bn.bn.running_var", "module.backbone_net3.sa4.mlp_module.layer2.conv.weight", "module.backbone_net3.sa4.mlp_module.layer2.bn.bn.weight", "module.backbone_net3.sa4.mlp_module.layer2.bn.bn.bias", "module.backbone_net3.sa4.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net3.sa4.mlp_module.layer2.bn.bn.running_var", "module.backbone_net3.fp1.mlp.layer0.conv.weight", "module.backbone_net3.fp1.mlp.layer0.bn.bn.weight", "module.backbone_net3.fp1.mlp.layer0.bn.bn.bias", "module.backbone_net3.fp1.mlp.layer0.bn.bn.running_mean", "module.backbone_net3.fp1.mlp.layer0.bn.bn.running_var", "module.backbone_net3.fp1.mlp.layer1.conv.weight", "module.backbone_net3.fp1.mlp.layer1.bn.bn.weight", "module.backbone_net3.fp1.mlp.layer1.bn.bn.bias", "module.backbone_net3.fp1.mlp.layer1.bn.bn.running_mean", "module.backbone_net3.fp1.mlp.layer1.bn.bn.running_var", "module.backbone_net3.fp2.mlp.layer0.conv.weight", "module.backbone_net3.fp2.mlp.layer0.bn.bn.weight", "module.backbone_net3.fp2.mlp.layer0.bn.bn.bias", "module.backbone_net3.fp2.mlp.layer0.bn.bn.running_mean", "module.backbone_net3.fp2.mlp.layer0.bn.bn.running_var", "module.backbone_net3.fp2.mlp.layer1.conv.weight", "module.backbone_net3.fp2.mlp.layer1.bn.bn.weight", "module.backbone_net3.fp2.mlp.layer1.bn.bn.bias", "module.backbone_net3.fp2.mlp.layer1.bn.bn.running_mean", "module.backbone_net3.fp2.mlp.layer1.bn.bn.running_var", "module.backbone_net4.sa1.mlp_module.layer0.conv.weight", "module.backbone_net4.sa1.mlp_module.layer0.bn.bn.weight", "module.backbone_net4.sa1.mlp_module.layer0.bn.bn.bias", "module.backbone_net4.sa1.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net4.sa1.mlp_module.layer0.bn.bn.running_var", "module.backbone_net4.sa1.mlp_module.layer1.conv.weight", "module.backbone_net4.sa1.mlp_module.layer1.bn.bn.weight", "module.backbone_net4.sa1.mlp_module.layer1.bn.bn.bias", "module.backbone_net4.sa1.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net4.sa1.mlp_module.layer1.bn.bn.running_var", "module.backbone_net4.sa1.mlp_module.layer2.conv.weight", "module.backbone_net4.sa1.mlp_module.layer2.bn.bn.weight", "module.backbone_net4.sa1.mlp_module.layer2.bn.bn.bias", "module.backbone_net4.sa1.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net4.sa1.mlp_module.layer2.bn.bn.running_var", "module.backbone_net4.sa2.mlp_module.layer0.conv.weight", "module.backbone_net4.sa2.mlp_module.layer0.bn.bn.weight", "module.backbone_net4.sa2.mlp_module.layer0.bn.bn.bias", "module.backbone_net4.sa2.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net4.sa2.mlp_module.layer0.bn.bn.running_var", "module.backbone_net4.sa2.mlp_module.layer1.conv.weight", "module.backbone_net4.sa2.mlp_module.layer1.bn.bn.weight", "module.backbone_net4.sa2.mlp_module.layer1.bn.bn.bias", "module.backbone_net4.sa2.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net4.sa2.mlp_module.layer1.bn.bn.running_var", "module.backbone_net4.sa2.mlp_module.layer2.conv.weight", "module.backbone_net4.sa2.mlp_module.layer2.bn.bn.weight", "module.backbone_net4.sa2.mlp_module.layer2.bn.bn.bias", "module.backbone_net4.sa2.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net4.sa2.mlp_module.layer2.bn.bn.running_var", "module.backbone_net4.sa3.mlp_module.layer0.conv.weight", "module.backbone_net4.sa3.mlp_module.layer0.bn.bn.weight", "module.backbone_net4.sa3.mlp_module.layer0.bn.bn.bias", "module.backbone_net4.sa3.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net4.sa3.mlp_module.layer0.bn.bn.running_var", "module.backbone_net4.sa3.mlp_module.layer1.conv.weight", "module.backbone_net4.sa3.mlp_module.layer1.bn.bn.weight", "module.backbone_net4.sa3.mlp_module.layer1.bn.bn.bias", "module.backbone_net4.sa3.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net4.sa3.mlp_module.layer1.bn.bn.running_var", "module.backbone_net4.sa3.mlp_module.layer2.conv.weight", "module.backbone_net4.sa3.mlp_module.layer2.bn.bn.weight", "module.backbone_net4.sa3.mlp_module.layer2.bn.bn.bias", "module.backbone_net4.sa3.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net4.sa3.mlp_module.layer2.bn.bn.running_var", "module.backbone_net4.sa4.mlp_module.layer0.conv.weight", "module.backbone_net4.sa4.mlp_module.layer0.bn.bn.weight", "module.backbone_net4.sa4.mlp_module.layer0.bn.bn.bias", "module.backbone_net4.sa4.mlp_module.layer0.bn.bn.running_mean", "module.backbone_net4.sa4.mlp_module.layer0.bn.bn.running_var", "module.backbone_net4.sa4.mlp_module.layer1.conv.weight", "module.backbone_net4.sa4.mlp_module.layer1.bn.bn.weight", "module.backbone_net4.sa4.mlp_module.layer1.bn.bn.bias", "module.backbone_net4.sa4.mlp_module.layer1.bn.bn.running_mean", "module.backbone_net4.sa4.mlp_module.layer1.bn.bn.running_var", "module.backbone_net4.sa4.mlp_module.layer2.conv.weight", "module.backbone_net4.sa4.mlp_module.layer2.bn.bn.weight", "module.backbone_net4.sa4.mlp_module.layer2.bn.bn.bias", "module.backbone_net4.sa4.mlp_module.layer2.bn.bn.running_mean", "module.backbone_net4.sa4.mlp_module.layer2.bn.bn.running_var", "module.backbone_net4.fp1.mlp.layer0.conv.weight", "module.backbone_net4.fp1.mlp.layer0.bn.bn.weight", "module.backbone_net4.fp1.mlp.layer0.bn.bn.bias", "module.backbone_net4.fp1.mlp.layer0.bn.bn.running_mean", "module.backbone_net4.fp1.mlp.layer0.bn.bn.running_var", "module.backbone_net4.fp1.mlp.layer1.conv.weight", "module.backbone_net4.fp1.mlp.layer1.bn.bn.weight", "module.backbone_net4.fp1.mlp.layer1.bn.bn.bias", "module.backbone_net4.fp1.mlp.layer1.bn.bn.running_mean", "module.backbone_net4.fp1.mlp.layer1.bn.bn.running_var", "module.backbone_net4.fp2.mlp.layer0.conv.weight", "module.backbone_net4.fp2.mlp.layer0.bn.bn.weight", "module.backbone_net4.fp2.mlp.layer0.bn.bn.bias", "module.backbone_net4.fp2.mlp.layer0.bn.bn.running_mean", "module.backbone_net4.fp2.mlp.layer0.bn.bn.running_var", "module.backbone_net4.fp2.mlp.layer1.conv.weight", "module.backbone_net4.fp2.mlp.layer1.bn.bn.weight", "module.backbone_net4.fp2.mlp.layer1.bn.bn.bias", "module.backbone_net4.fp2.mlp.layer1.bn.bn.running_mean", "module.backbone_net4.fp2.mlp.layer1.bn.bn.running_var", "module.conv_agg1.weight", "module.conv_agg1.bias", "module.bn_agg1.weight", "module.bn_agg1.bias", "module.bn_agg1.running_mean", "module.bn_agg1.running_var", "module.conv_agg2.weight", "module.conv_agg2.bias", "module.bn_agg2.weight", "module.bn_agg2.bias", "module.bn_agg2.running_mean", "module.bn_agg2.running_var", "module.conv_flag_z1.weight", "module.conv_flag_z1.bias", "module.bn_flag_z1.weight", "module.bn_flag_z1.bias", "module.bn_flag_z1.running_mean", "module.bn_flag_z1.running_var", "module.conv_flag_z2.weight", "module.conv_flag_z2.bias", "module.conv_flag_xy1.weight", "module.conv_flag_xy1.bias", "module.bn_flag_xy1.weight", "module.bn_flag_xy1.bias", "module.bn_flag_xy1.running_mean", "module.bn_flag_xy1.running_var", "module.conv_flag_xy2.weight", "module.conv_flag_xy2.bias", "module.conv_flag_line1.weight", "module.conv_flag_line1.bias", "module.bn_flag_line1.weight", "module.bn_flag_line1.bias", "module.bn_flag_line1.running_mean", "module.bn_flag_line1.running_var", "module.conv_flag_line2.weight", "module.conv_flag_line2.bias", "module.vgen.conv1.weight", "module.vgen.conv1.bias", "module.vgen.conv2.weight", "module.vgen.conv2.bias", "module.vgen.conv3.weight", "module.vgen.conv3.bias", "module.vgen.bn1.weight", "module.vgen.bn1.bias", "module.vgen.bn1.running_mean", "module.vgen.bn1.running_var", "module.vgen.bn2.weight", "module.vgen.bn2.bias", "module.vgen.bn2.running_mean", "module.vgen.bn2.running_var", "module.vgen_z.conv1.weight", "module.vgen_z.conv1.bias", "module.vgen_z.conv2.weight", "module.vgen_z.conv2.bias", "module.vgen_z.conv3.weight", "module.vgen_z.conv3.bias", "module.vgen_z.bn1.weight", "module.vgen_z.bn1.bias", "module.vgen_z.bn1.running_mean", "module.vgen_z.bn1.running_var", "module.vgen_z.bn2.weight", "module.vgen_z.bn2.bias", "module.vgen_z.bn2.running_mean", "module.vgen_z.bn2.running_var", "module.vgen_xy.conv1.weight", "module.vgen_xy.conv1.bias", "module.vgen_xy.conv2.weight", "module.vgen_xy.conv2.bias", "module.vgen_xy.conv3.weight", "module.vgen_xy.conv3.bias", "module.vgen_xy.bn1.weight", "module.vgen_xy.bn1.bias", "module.vgen_xy.bn1.running_mean", "module.vgen_xy.bn1.running_var", "module.vgen_xy.bn2.weight", "module.vgen_xy.bn2.bias", "module.vgen_xy.bn2.running_mean", "module.vgen_xy.bn2.running_var", "module.vgen_line.conv1.weight", "module.vgen_line.conv1.bias", "module.vgen_line.conv2.weight", "module.vgen_line.conv2.bias", "module.vgen_line.conv3.weight", "module.vgen_line.conv3.bias", "module.vgen_line.bn1.weight", "module.vgen_line.bn1.bias", "module.vgen_line.bn1.running_mean", "module.vgen_line.bn1.running_var", "module.vgen_line.bn2.weight", "module.vgen_line.bn2.bias", "module.vgen_line.bn2.running_mean", "module.vgen_line.bn2.running_var", "module.pnet_z.vote_aggregation.mlp_module.layer0.conv.weight", "module.pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.weight", "module.pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.bias", "module.pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "module.pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "module.pnet_z.vote_aggregation.mlp_module.layer1.conv.weight", "module.pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.weight", "module.pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.bias", "module.pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "module.pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "module.pnet_z.vote_aggregation.mlp_module.layer2.conv.weight", "module.pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.weight", "module.pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.bias", "module.pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "module.pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "module.pnet_z.conv1.weight", "module.pnet_z.conv1.bias", "module.pnet_z.conv2.weight", "module.pnet_z.conv2.bias", "module.pnet_z.conv3.weight", "module.pnet_z.conv3.bias", "module.pnet_z.bn1.weight", "module.pnet_z.bn1.bias", "module.pnet_z.bn1.running_mean", "module.pnet_z.bn1.running_var", "module.pnet_z.bn2.weight", "module.pnet_z.bn2.bias", "module.pnet_z.bn2.running_mean", "module.pnet_z.bn2.running_var", "module.pnet_xy.vote_aggregation.mlp_module.layer0.conv.weight", "module.pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.weight", "module.pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.bias", "module.pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "module.pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "module.pnet_xy.vote_aggregation.mlp_module.layer1.conv.weight", "module.pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.weight", "module.pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.bias", "module.pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "module.pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "module.pnet_xy.vote_aggregation.mlp_module.layer2.conv.weight", "module.pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.weight", "module.pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.bias", "module.pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "module.pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "module.pnet_xy.conv1.weight", "module.pnet_xy.conv1.bias", "module.pnet_xy.conv2.weight", "module.pnet_xy.conv2.bias", "module.pnet_xy.conv3.weight", "module.pnet_xy.conv3.bias", "module.pnet_xy.bn1.weight", "module.pnet_xy.bn1.bias", "module.pnet_xy.bn1.running_mean", "module.pnet_xy.bn1.running_var", "module.pnet_xy.bn2.weight", "module.pnet_xy.bn2.bias", "module.pnet_xy.bn2.running_mean", "module.pnet_xy.bn2.running_var", "module.pnet_line.vote_aggregation.mlp_module.layer0.conv.weight", "module.pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.weight", "module.pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.bias", "module.pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "module.pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "module.pnet_line.vote_aggregation.mlp_module.layer1.conv.weight", "module.pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.weight", "module.pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.bias", "module.pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "module.pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "module.pnet_line.vote_aggregation.mlp_module.layer2.conv.weight", "module.pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.weight", "module.pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.bias", "module.pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "module.pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "module.pnet_line.conv1.weight", "module.pnet_line.conv1.bias", "module.pnet_line.conv2.weight", "module.pnet_line.conv2.bias", "module.pnet_line.conv3.weight", "module.pnet_line.conv3.bias", "module.pnet_line.bn1.weight", "module.pnet_line.bn1.bias", "module.pnet_line.bn1.running_mean", "module.pnet_line.bn1.running_var", "module.pnet_line.bn2.weight", "module.pnet_line.bn2.bias", "module.pnet_line.bn2.running_mean", "module.pnet_line.bn2.running_var", "module.pnet_final.vote_aggregation.mlp_module.layer0.conv.weight", "module.pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.weight", "module.pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.bias", "module.pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "module.pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "module.pnet_final.vote_aggregation.mlp_module.layer1.conv.weight", "module.pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.weight", "module.pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.bias", "module.pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "module.pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "module.pnet_final.vote_aggregation.mlp_module.layer2.conv.weight", "module.pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.weight", "module.pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.bias", "module.pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "module.pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "module.pnet_final.match_surface_center.mlp_module.layer0.conv.weight", "module.pnet_final.match_surface_center.mlp_module.layer0.bn.bn.weight", "module.pnet_final.match_surface_center.mlp_module.layer0.bn.bn.bias", "module.pnet_final.match_surface_center.mlp_module.layer0.bn.bn.running_mean", "module.pnet_final.match_surface_center.mlp_module.layer0.bn.bn.running_var", "module.pnet_final.match_surface_center.mlp_module.layer1.conv.weight", "module.pnet_final.match_surface_center.mlp_module.layer1.bn.bn.weight", "module.pnet_final.match_surface_center.mlp_module.layer1.bn.bn.bias", "module.pnet_final.match_surface_center.mlp_module.layer1.bn.bn.running_mean", "module.pnet_final.match_surface_center.mlp_module.layer1.bn.bn.running_var", "module.pnet_final.match_surface_center.mlp_module.layer2.conv.weight", "module.pnet_final.match_surface_center.mlp_module.layer2.bn.bn.weight", "module.pnet_final.match_surface_center.mlp_module.layer2.bn.bn.bias", "module.pnet_final.match_surface_center.mlp_module.layer2.bn.bn.running_mean", "module.pnet_final.match_surface_center.mlp_module.layer2.bn.bn.running_var", "module.pnet_final.match_line_center.mlp_module.layer0.conv.weight", "module.pnet_final.match_line_center.mlp_module.layer0.bn.bn.weight", "module.pnet_final.match_line_center.mlp_module.layer0.bn.bn.bias", "module.pnet_final.match_line_center.mlp_module.layer0.bn.bn.running_mean", "module.pnet_final.match_line_center.mlp_module.layer0.bn.bn.running_var", "module.pnet_final.match_line_center.mlp_module.layer1.conv.weight", "module.pnet_final.match_line_center.mlp_module.layer1.bn.bn.weight", "module.pnet_final.match_line_center.mlp_module.layer1.bn.bn.bias", "module.pnet_final.match_line_center.mlp_module.layer1.bn.bn.running_mean", "module.pnet_final.match_line_center.mlp_module.layer1.bn.bn.running_var", "module.pnet_final.match_line_center.mlp_module.layer2.conv.weight", "module.pnet_final.match_line_center.mlp_module.layer2.bn.bn.weight", "module.pnet_final.match_line_center.mlp_module.layer2.bn.bn.bias", "module.pnet_final.match_line_center.mlp_module.layer2.bn.bn.running_mean", "module.pnet_final.match_line_center.mlp_module.layer2.bn.bn.running_var", "module.pnet_final.conv1.weight", "module.pnet_final.conv1.bias", "module.pnet_final.conv2.weight", "module.pnet_final.conv2.bias", "module.pnet_final.conv3.weight", "module.pnet_final.conv3.bias", "module.pnet_final.bn1.weight", "module.pnet_final.bn1.bias", "module.pnet_final.bn1.running_mean", "module.pnet_final.bn1.running_var", "module.pnet_final.bn2.weight", "module.pnet_final.bn2.bias", "module.pnet_final.bn2.running_mean", "module.pnet_final.bn2.running_var", "module.pnet_final.conv_match1.weight", "module.pnet_final.conv_match1.bias", "module.pnet_final.conv_match2.weight", "module.pnet_final.conv_match2.bias", "module.pnet_final.bn_match1.weight", "module.pnet_final.bn_match1.bias", "module.pnet_final.bn_match1.running_mean", "module.pnet_final.bn_match1.running_var", "module.pnet_final.conv_match_sem1.weight", "module.pnet_final.conv_match_sem1.bias", "module.pnet_final.conv_match_sem2.weight", "module.pnet_final.conv_match_sem2.bias", "module.pnet_final.bn_match_sem1.weight", "module.pnet_final.bn_match_sem1.bias", "module.pnet_final.bn_match_sem1.running_mean", "module.pnet_final.bn_match_sem1.running_var", "module.pnet_final.conv_surface1.weight", "module.pnet_final.conv_surface1.bias", "module.pnet_final.conv_surface2.weight", "module.pnet_final.conv_surface2.bias", "module.pnet_final.bn_surface1.weight", "module.pnet_final.bn_surface1.bias", "module.pnet_final.bn_surface1.running_mean", "module.pnet_final.bn_surface1.running_var", "module.pnet_final.bn_surface2.weight", "module.pnet_final.bn_surface2.bias", "module.pnet_final.bn_surface2.running_mean", "module.pnet_final.bn_surface2.running_var", "module.pnet_final.conv_line1.weight", "module.pnet_final.conv_line1.bias", "module.pnet_final.conv_line2.weight", "module.pnet_final.conv_line2.bias", "module.pnet_final.bn_line1.weight", "module.pnet_final.bn_line1.bias", "module.pnet_final.bn_line1.running_mean", "module.pnet_final.bn_line1.running_var", "module.pnet_final.bn_line2.weight", "module.pnet_final.bn_line2.bias", "module.pnet_final.bn_line2.running_mean", "module.pnet_final.bn_line2.running_var", "module.pnet_final.conv_refine1.weight", "module.pnet_final.conv_refine1.bias", "module.pnet_final.conv_refine2.weight", "module.pnet_final.conv_refine2.bias", "module.pnet_final.conv_refine3.weight", "module.pnet_final.conv_refine3.bias", "module.pnet_final.conv_refine4.weight", "module.pnet_final.conv_refine4.bias", "module.pnet_final.bn_refine1.weight", "module.pnet_final.bn_refine1.bias", "module.pnet_final.bn_refine1.running_mean", "module.pnet_final.bn_refine1.running_var", "module.pnet_final.bn_refine2.weight", "module.pnet_final.bn_refine2.bias", "module.pnet_final.bn_refine2.running_mean", "module.pnet_final.bn_refine2.running_var", "module.pnet_final.bn_refine3.weight", "module.pnet_final.bn_refine3.bias", "module.pnet_final.bn_refine3.running_mean", "module.pnet_final.bn_refine3.running_var". Unexpected key(s) in state_dict: "backbone_net1.sa1.mlp_module.layer0.conv.weight", "backbone_net1.sa1.mlp_module.layer0.bn.bn.weight", "backbone_net1.sa1.mlp_module.layer0.bn.bn.bias", "backbone_net1.sa1.mlp_module.layer0.bn.bn.running_mean", "backbone_net1.sa1.mlp_module.layer0.bn.bn.running_var", "backbone_net1.sa1.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net1.sa1.mlp_module.layer1.conv.weight", "backbone_net1.sa1.mlp_module.layer1.bn.bn.weight", "backbone_net1.sa1.mlp_module.layer1.bn.bn.bias", "backbone_net1.sa1.mlp_module.layer1.bn.bn.running_mean", "backbone_net1.sa1.mlp_module.layer1.bn.bn.running_var", "backbone_net1.sa1.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net1.sa1.mlp_module.layer2.conv.weight", "backbone_net1.sa1.mlp_module.layer2.bn.bn.weight", "backbone_net1.sa1.mlp_module.layer2.bn.bn.bias", "backbone_net1.sa1.mlp_module.layer2.bn.bn.running_mean", "backbone_net1.sa1.mlp_module.layer2.bn.bn.running_var", "backbone_net1.sa1.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net1.sa2.mlp_module.layer0.conv.weight", "backbone_net1.sa2.mlp_module.layer0.bn.bn.weight", "backbone_net1.sa2.mlp_module.layer0.bn.bn.bias", "backbone_net1.sa2.mlp_module.layer0.bn.bn.running_mean", "backbone_net1.sa2.mlp_module.layer0.bn.bn.running_var", "backbone_net1.sa2.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net1.sa2.mlp_module.layer1.conv.weight", "backbone_net1.sa2.mlp_module.layer1.bn.bn.weight", "backbone_net1.sa2.mlp_module.layer1.bn.bn.bias", "backbone_net1.sa2.mlp_module.layer1.bn.bn.running_mean", "backbone_net1.sa2.mlp_module.layer1.bn.bn.running_var", "backbone_net1.sa2.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net1.sa2.mlp_module.layer2.conv.weight", "backbone_net1.sa2.mlp_module.layer2.bn.bn.weight", "backbone_net1.sa2.mlp_module.layer2.bn.bn.bias", "backbone_net1.sa2.mlp_module.layer2.bn.bn.running_mean", "backbone_net1.sa2.mlp_module.layer2.bn.bn.running_var", "backbone_net1.sa2.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net1.sa3.mlp_module.layer0.conv.weight", "backbone_net1.sa3.mlp_module.layer0.bn.bn.weight", "backbone_net1.sa3.mlp_module.layer0.bn.bn.bias", "backbone_net1.sa3.mlp_module.layer0.bn.bn.running_mean", "backbone_net1.sa3.mlp_module.layer0.bn.bn.running_var", "backbone_net1.sa3.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net1.sa3.mlp_module.layer1.conv.weight", "backbone_net1.sa3.mlp_module.layer1.bn.bn.weight", "backbone_net1.sa3.mlp_module.layer1.bn.bn.bias", "backbone_net1.sa3.mlp_module.layer1.bn.bn.running_mean", "backbone_net1.sa3.mlp_module.layer1.bn.bn.running_var", "backbone_net1.sa3.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net1.sa3.mlp_module.layer2.conv.weight", "backbone_net1.sa3.mlp_module.layer2.bn.bn.weight", "backbone_net1.sa3.mlp_module.layer2.bn.bn.bias", "backbone_net1.sa3.mlp_module.layer2.bn.bn.running_mean", "backbone_net1.sa3.mlp_module.layer2.bn.bn.running_var", "backbone_net1.sa3.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net1.sa4.mlp_module.layer0.conv.weight", "backbone_net1.sa4.mlp_module.layer0.bn.bn.weight", "backbone_net1.sa4.mlp_module.layer0.bn.bn.bias", "backbone_net1.sa4.mlp_module.layer0.bn.bn.running_mean", "backbone_net1.sa4.mlp_module.layer0.bn.bn.running_var", "backbone_net1.sa4.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net1.sa4.mlp_module.layer1.conv.weight", "backbone_net1.sa4.mlp_module.layer1.bn.bn.weight", "backbone_net1.sa4.mlp_module.layer1.bn.bn.bias", "backbone_net1.sa4.mlp_module.layer1.bn.bn.running_mean", "backbone_net1.sa4.mlp_module.layer1.bn.bn.running_var", "backbone_net1.sa4.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net1.sa4.mlp_module.layer2.conv.weight", "backbone_net1.sa4.mlp_module.layer2.bn.bn.weight", "backbone_net1.sa4.mlp_module.layer2.bn.bn.bias", "backbone_net1.sa4.mlp_module.layer2.bn.bn.running_mean", "backbone_net1.sa4.mlp_module.layer2.bn.bn.running_var", "backbone_net1.sa4.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net1.fp1.mlp.layer0.conv.weight", "backbone_net1.fp1.mlp.layer0.bn.bn.weight", "backbone_net1.fp1.mlp.layer0.bn.bn.bias", "backbone_net1.fp1.mlp.layer0.bn.bn.running_mean", "backbone_net1.fp1.mlp.layer0.bn.bn.running_var", "backbone_net1.fp1.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net1.fp1.mlp.layer1.conv.weight", "backbone_net1.fp1.mlp.layer1.bn.bn.weight", "backbone_net1.fp1.mlp.layer1.bn.bn.bias", "backbone_net1.fp1.mlp.layer1.bn.bn.running_mean", "backbone_net1.fp1.mlp.layer1.bn.bn.running_var", "backbone_net1.fp1.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net1.fp2.mlp.layer0.conv.weight", "backbone_net1.fp2.mlp.layer0.bn.bn.weight", "backbone_net1.fp2.mlp.layer0.bn.bn.bias", "backbone_net1.fp2.mlp.layer0.bn.bn.running_mean", "backbone_net1.fp2.mlp.layer0.bn.bn.running_var", "backbone_net1.fp2.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net1.fp2.mlp.layer1.conv.weight", "backbone_net1.fp2.mlp.layer1.bn.bn.weight", "backbone_net1.fp2.mlp.layer1.bn.bn.bias", "backbone_net1.fp2.mlp.layer1.bn.bn.running_mean", "backbone_net1.fp2.mlp.layer1.bn.bn.running_var", "backbone_net1.fp2.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net2.sa1.mlp_module.layer0.conv.weight", "backbone_net2.sa1.mlp_module.layer0.bn.bn.weight", "backbone_net2.sa1.mlp_module.layer0.bn.bn.bias", "backbone_net2.sa1.mlp_module.layer0.bn.bn.running_mean", "backbone_net2.sa1.mlp_module.layer0.bn.bn.running_var", "backbone_net2.sa1.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net2.sa1.mlp_module.layer1.conv.weight", "backbone_net2.sa1.mlp_module.layer1.bn.bn.weight", "backbone_net2.sa1.mlp_module.layer1.bn.bn.bias", "backbone_net2.sa1.mlp_module.layer1.bn.bn.running_mean", "backbone_net2.sa1.mlp_module.layer1.bn.bn.running_var", "backbone_net2.sa1.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net2.sa1.mlp_module.layer2.conv.weight", "backbone_net2.sa1.mlp_module.layer2.bn.bn.weight", "backbone_net2.sa1.mlp_module.layer2.bn.bn.bias", "backbone_net2.sa1.mlp_module.layer2.bn.bn.running_mean", "backbone_net2.sa1.mlp_module.layer2.bn.bn.running_var", "backbone_net2.sa1.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net2.sa2.mlp_module.layer0.conv.weight", "backbone_net2.sa2.mlp_module.layer0.bn.bn.weight", "backbone_net2.sa2.mlp_module.layer0.bn.bn.bias", "backbone_net2.sa2.mlp_module.layer0.bn.bn.running_mean", "backbone_net2.sa2.mlp_module.layer0.bn.bn.running_var", "backbone_net2.sa2.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net2.sa2.mlp_module.layer1.conv.weight", "backbone_net2.sa2.mlp_module.layer1.bn.bn.weight", "backbone_net2.sa2.mlp_module.layer1.bn.bn.bias", "backbone_net2.sa2.mlp_module.layer1.bn.bn.running_mean", "backbone_net2.sa2.mlp_module.layer1.bn.bn.running_var", "backbone_net2.sa2.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net2.sa2.mlp_module.layer2.conv.weight", "backbone_net2.sa2.mlp_module.layer2.bn.bn.weight", "backbone_net2.sa2.mlp_module.layer2.bn.bn.bias", "backbone_net2.sa2.mlp_module.layer2.bn.bn.running_mean", "backbone_net2.sa2.mlp_module.layer2.bn.bn.running_var", "backbone_net2.sa2.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net2.sa3.mlp_module.layer0.conv.weight", "backbone_net2.sa3.mlp_module.layer0.bn.bn.weight", "backbone_net2.sa3.mlp_module.layer0.bn.bn.bias", "backbone_net2.sa3.mlp_module.layer0.bn.bn.running_mean", "backbone_net2.sa3.mlp_module.layer0.bn.bn.running_var", "backbone_net2.sa3.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net2.sa3.mlp_module.layer1.conv.weight", "backbone_net2.sa3.mlp_module.layer1.bn.bn.weight", "backbone_net2.sa3.mlp_module.layer1.bn.bn.bias", "backbone_net2.sa3.mlp_module.layer1.bn.bn.running_mean", "backbone_net2.sa3.mlp_module.layer1.bn.bn.running_var", "backbone_net2.sa3.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net2.sa3.mlp_module.layer2.conv.weight", "backbone_net2.sa3.mlp_module.layer2.bn.bn.weight", "backbone_net2.sa3.mlp_module.layer2.bn.bn.bias", "backbone_net2.sa3.mlp_module.layer2.bn.bn.running_mean", "backbone_net2.sa3.mlp_module.layer2.bn.bn.running_var", "backbone_net2.sa3.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net2.sa4.mlp_module.layer0.conv.weight", "backbone_net2.sa4.mlp_module.layer0.bn.bn.weight", "backbone_net2.sa4.mlp_module.layer0.bn.bn.bias", "backbone_net2.sa4.mlp_module.layer0.bn.bn.running_mean", "backbone_net2.sa4.mlp_module.layer0.bn.bn.running_var", "backbone_net2.sa4.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net2.sa4.mlp_module.layer1.conv.weight", "backbone_net2.sa4.mlp_module.layer1.bn.bn.weight", "backbone_net2.sa4.mlp_module.layer1.bn.bn.bias", "backbone_net2.sa4.mlp_module.layer1.bn.bn.running_mean", "backbone_net2.sa4.mlp_module.layer1.bn.bn.running_var", "backbone_net2.sa4.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net2.sa4.mlp_module.layer2.conv.weight", "backbone_net2.sa4.mlp_module.layer2.bn.bn.weight", "backbone_net2.sa4.mlp_module.layer2.bn.bn.bias", "backbone_net2.sa4.mlp_module.layer2.bn.bn.running_mean", "backbone_net2.sa4.mlp_module.layer2.bn.bn.running_var", "backbone_net2.sa4.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net2.fp1.mlp.layer0.conv.weight", "backbone_net2.fp1.mlp.layer0.bn.bn.weight", "backbone_net2.fp1.mlp.layer0.bn.bn.bias", "backbone_net2.fp1.mlp.layer0.bn.bn.running_mean", "backbone_net2.fp1.mlp.layer0.bn.bn.running_var", "backbone_net2.fp1.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net2.fp1.mlp.layer1.conv.weight", "backbone_net2.fp1.mlp.layer1.bn.bn.weight", "backbone_net2.fp1.mlp.layer1.bn.bn.bias", "backbone_net2.fp1.mlp.layer1.bn.bn.running_mean", "backbone_net2.fp1.mlp.layer1.bn.bn.running_var", "backbone_net2.fp1.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net2.fp2.mlp.layer0.conv.weight", "backbone_net2.fp2.mlp.layer0.bn.bn.weight", "backbone_net2.fp2.mlp.layer0.bn.bn.bias", "backbone_net2.fp2.mlp.layer0.bn.bn.running_mean", "backbone_net2.fp2.mlp.layer0.bn.bn.running_var", "backbone_net2.fp2.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net2.fp2.mlp.layer1.conv.weight", "backbone_net2.fp2.mlp.layer1.bn.bn.weight", "backbone_net2.fp2.mlp.layer1.bn.bn.bias", "backbone_net2.fp2.mlp.layer1.bn.bn.running_mean", "backbone_net2.fp2.mlp.layer1.bn.bn.running_var", "backbone_net2.fp2.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net3.sa1.mlp_module.layer0.conv.weight", "backbone_net3.sa1.mlp_module.layer0.bn.bn.weight", "backbone_net3.sa1.mlp_module.layer0.bn.bn.bias", "backbone_net3.sa1.mlp_module.layer0.bn.bn.running_mean", "backbone_net3.sa1.mlp_module.layer0.bn.bn.running_var", "backbone_net3.sa1.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net3.sa1.mlp_module.layer1.conv.weight", "backbone_net3.sa1.mlp_module.layer1.bn.bn.weight", "backbone_net3.sa1.mlp_module.layer1.bn.bn.bias", "backbone_net3.sa1.mlp_module.layer1.bn.bn.running_mean", "backbone_net3.sa1.mlp_module.layer1.bn.bn.running_var", "backbone_net3.sa1.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net3.sa1.mlp_module.layer2.conv.weight", "backbone_net3.sa1.mlp_module.layer2.bn.bn.weight", "backbone_net3.sa1.mlp_module.layer2.bn.bn.bias", "backbone_net3.sa1.mlp_module.layer2.bn.bn.running_mean", "backbone_net3.sa1.mlp_module.layer2.bn.bn.running_var", "backbone_net3.sa1.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net3.sa2.mlp_module.layer0.conv.weight", "backbone_net3.sa2.mlp_module.layer0.bn.bn.weight", "backbone_net3.sa2.mlp_module.layer0.bn.bn.bias", "backbone_net3.sa2.mlp_module.layer0.bn.bn.running_mean", "backbone_net3.sa2.mlp_module.layer0.bn.bn.running_var", "backbone_net3.sa2.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net3.sa2.mlp_module.layer1.conv.weight", "backbone_net3.sa2.mlp_module.layer1.bn.bn.weight", "backbone_net3.sa2.mlp_module.layer1.bn.bn.bias", "backbone_net3.sa2.mlp_module.layer1.bn.bn.running_mean", "backbone_net3.sa2.mlp_module.layer1.bn.bn.running_var", "backbone_net3.sa2.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net3.sa2.mlp_module.layer2.conv.weight", "backbone_net3.sa2.mlp_module.layer2.bn.bn.weight", "backbone_net3.sa2.mlp_module.layer2.bn.bn.bias", "backbone_net3.sa2.mlp_module.layer2.bn.bn.running_mean", "backbone_net3.sa2.mlp_module.layer2.bn.bn.running_var", "backbone_net3.sa2.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net3.sa3.mlp_module.layer0.conv.weight", "backbone_net3.sa3.mlp_module.layer0.bn.bn.weight", "backbone_net3.sa3.mlp_module.layer0.bn.bn.bias", "backbone_net3.sa3.mlp_module.layer0.bn.bn.running_mean", "backbone_net3.sa3.mlp_module.layer0.bn.bn.running_var", "backbone_net3.sa3.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net3.sa3.mlp_module.layer1.conv.weight", "backbone_net3.sa3.mlp_module.layer1.bn.bn.weight", "backbone_net3.sa3.mlp_module.layer1.bn.bn.bias", "backbone_net3.sa3.mlp_module.layer1.bn.bn.running_mean", "backbone_net3.sa3.mlp_module.layer1.bn.bn.running_var", "backbone_net3.sa3.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net3.sa3.mlp_module.layer2.conv.weight", "backbone_net3.sa3.mlp_module.layer2.bn.bn.weight", "backbone_net3.sa3.mlp_module.layer2.bn.bn.bias", "backbone_net3.sa3.mlp_module.layer2.bn.bn.running_mean", "backbone_net3.sa3.mlp_module.layer2.bn.bn.running_var", "backbone_net3.sa3.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net3.sa4.mlp_module.layer0.conv.weight", "backbone_net3.sa4.mlp_module.layer0.bn.bn.weight", "backbone_net3.sa4.mlp_module.layer0.bn.bn.bias", "backbone_net3.sa4.mlp_module.layer0.bn.bn.running_mean", "backbone_net3.sa4.mlp_module.layer0.bn.bn.running_var", "backbone_net3.sa4.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net3.sa4.mlp_module.layer1.conv.weight", "backbone_net3.sa4.mlp_module.layer1.bn.bn.weight", "backbone_net3.sa4.mlp_module.layer1.bn.bn.bias", "backbone_net3.sa4.mlp_module.layer1.bn.bn.running_mean", "backbone_net3.sa4.mlp_module.layer1.bn.bn.running_var", "backbone_net3.sa4.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net3.sa4.mlp_module.layer2.conv.weight", "backbone_net3.sa4.mlp_module.layer2.bn.bn.weight", "backbone_net3.sa4.mlp_module.layer2.bn.bn.bias", "backbone_net3.sa4.mlp_module.layer2.bn.bn.running_mean", "backbone_net3.sa4.mlp_module.layer2.bn.bn.running_var", "backbone_net3.sa4.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net3.fp1.mlp.layer0.conv.weight", "backbone_net3.fp1.mlp.layer0.bn.bn.weight", "backbone_net3.fp1.mlp.layer0.bn.bn.bias", "backbone_net3.fp1.mlp.layer0.bn.bn.running_mean", "backbone_net3.fp1.mlp.layer0.bn.bn.running_var", "backbone_net3.fp1.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net3.fp1.mlp.layer1.conv.weight", "backbone_net3.fp1.mlp.layer1.bn.bn.weight", "backbone_net3.fp1.mlp.layer1.bn.bn.bias", "backbone_net3.fp1.mlp.layer1.bn.bn.running_mean", "backbone_net3.fp1.mlp.layer1.bn.bn.running_var", "backbone_net3.fp1.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net3.fp2.mlp.layer0.conv.weight", "backbone_net3.fp2.mlp.layer0.bn.bn.weight", "backbone_net3.fp2.mlp.layer0.bn.bn.bias", "backbone_net3.fp2.mlp.layer0.bn.bn.running_mean", "backbone_net3.fp2.mlp.layer0.bn.bn.running_var", "backbone_net3.fp2.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net3.fp2.mlp.layer1.conv.weight", "backbone_net3.fp2.mlp.layer1.bn.bn.weight", "backbone_net3.fp2.mlp.layer1.bn.bn.bias", "backbone_net3.fp2.mlp.layer1.bn.bn.running_mean", "backbone_net3.fp2.mlp.layer1.bn.bn.running_var", "backbone_net3.fp2.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net4.sa1.mlp_module.layer0.conv.weight", "backbone_net4.sa1.mlp_module.layer0.bn.bn.weight", "backbone_net4.sa1.mlp_module.layer0.bn.bn.bias", "backbone_net4.sa1.mlp_module.layer0.bn.bn.running_mean", "backbone_net4.sa1.mlp_module.layer0.bn.bn.running_var", "backbone_net4.sa1.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net4.sa1.mlp_module.layer1.conv.weight", "backbone_net4.sa1.mlp_module.layer1.bn.bn.weight", "backbone_net4.sa1.mlp_module.layer1.bn.bn.bias", "backbone_net4.sa1.mlp_module.layer1.bn.bn.running_mean", "backbone_net4.sa1.mlp_module.layer1.bn.bn.running_var", "backbone_net4.sa1.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net4.sa1.mlp_module.layer2.conv.weight", "backbone_net4.sa1.mlp_module.layer2.bn.bn.weight", "backbone_net4.sa1.mlp_module.layer2.bn.bn.bias", "backbone_net4.sa1.mlp_module.layer2.bn.bn.running_mean", "backbone_net4.sa1.mlp_module.layer2.bn.bn.running_var", "backbone_net4.sa1.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net4.sa2.mlp_module.layer0.conv.weight", "backbone_net4.sa2.mlp_module.layer0.bn.bn.weight", "backbone_net4.sa2.mlp_module.layer0.bn.bn.bias", "backbone_net4.sa2.mlp_module.layer0.bn.bn.running_mean", "backbone_net4.sa2.mlp_module.layer0.bn.bn.running_var", "backbone_net4.sa2.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net4.sa2.mlp_module.layer1.conv.weight", "backbone_net4.sa2.mlp_module.layer1.bn.bn.weight", "backbone_net4.sa2.mlp_module.layer1.bn.bn.bias", "backbone_net4.sa2.mlp_module.layer1.bn.bn.running_mean", "backbone_net4.sa2.mlp_module.layer1.bn.bn.running_var", "backbone_net4.sa2.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net4.sa2.mlp_module.layer2.conv.weight", "backbone_net4.sa2.mlp_module.layer2.bn.bn.weight", "backbone_net4.sa2.mlp_module.layer2.bn.bn.bias", "backbone_net4.sa2.mlp_module.layer2.bn.bn.running_mean", "backbone_net4.sa2.mlp_module.layer2.bn.bn.running_var", "backbone_net4.sa2.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net4.sa3.mlp_module.layer0.conv.weight", "backbone_net4.sa3.mlp_module.layer0.bn.bn.weight", "backbone_net4.sa3.mlp_module.layer0.bn.bn.bias", "backbone_net4.sa3.mlp_module.layer0.bn.bn.running_mean", "backbone_net4.sa3.mlp_module.layer0.bn.bn.running_var", "backbone_net4.sa3.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net4.sa3.mlp_module.layer1.conv.weight", "backbone_net4.sa3.mlp_module.layer1.bn.bn.weight", "backbone_net4.sa3.mlp_module.layer1.bn.bn.bias", "backbone_net4.sa3.mlp_module.layer1.bn.bn.running_mean", "backbone_net4.sa3.mlp_module.layer1.bn.bn.running_var", "backbone_net4.sa3.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net4.sa3.mlp_module.layer2.conv.weight", "backbone_net4.sa3.mlp_module.layer2.bn.bn.weight", "backbone_net4.sa3.mlp_module.layer2.bn.bn.bias", "backbone_net4.sa3.mlp_module.layer2.bn.bn.running_mean", "backbone_net4.sa3.mlp_module.layer2.bn.bn.running_var", "backbone_net4.sa3.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net4.sa4.mlp_module.layer0.conv.weight", "backbone_net4.sa4.mlp_module.layer0.bn.bn.weight", "backbone_net4.sa4.mlp_module.layer0.bn.bn.bias", "backbone_net4.sa4.mlp_module.layer0.bn.bn.running_mean", "backbone_net4.sa4.mlp_module.layer0.bn.bn.running_var", "backbone_net4.sa4.mlp_module.layer0.bn.bn.num_batches_tracked", "backbone_net4.sa4.mlp_module.layer1.conv.weight", "backbone_net4.sa4.mlp_module.layer1.bn.bn.weight", "backbone_net4.sa4.mlp_module.layer1.bn.bn.bias", "backbone_net4.sa4.mlp_module.layer1.bn.bn.running_mean", "backbone_net4.sa4.mlp_module.layer1.bn.bn.running_var", "backbone_net4.sa4.mlp_module.layer1.bn.bn.num_batches_tracked", "backbone_net4.sa4.mlp_module.layer2.conv.weight", "backbone_net4.sa4.mlp_module.layer2.bn.bn.weight", "backbone_net4.sa4.mlp_module.layer2.bn.bn.bias", "backbone_net4.sa4.mlp_module.layer2.bn.bn.running_mean", "backbone_net4.sa4.mlp_module.layer2.bn.bn.running_var", "backbone_net4.sa4.mlp_module.layer2.bn.bn.num_batches_tracked", "backbone_net4.fp1.mlp.layer0.conv.weight", "backbone_net4.fp1.mlp.layer0.bn.bn.weight", "backbone_net4.fp1.mlp.layer0.bn.bn.bias", "backbone_net4.fp1.mlp.layer0.bn.bn.running_mean", "backbone_net4.fp1.mlp.layer0.bn.bn.running_var", "backbone_net4.fp1.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net4.fp1.mlp.layer1.conv.weight", "backbone_net4.fp1.mlp.layer1.bn.bn.weight", "backbone_net4.fp1.mlp.layer1.bn.bn.bias", "backbone_net4.fp1.mlp.layer1.bn.bn.running_mean", "backbone_net4.fp1.mlp.layer1.bn.bn.running_var", "backbone_net4.fp1.mlp.layer1.bn.bn.num_batches_tracked", "backbone_net4.fp2.mlp.layer0.conv.weight", "backbone_net4.fp2.mlp.layer0.bn.bn.weight", "backbone_net4.fp2.mlp.layer0.bn.bn.bias", "backbone_net4.fp2.mlp.layer0.bn.bn.running_mean", "backbone_net4.fp2.mlp.layer0.bn.bn.running_var", "backbone_net4.fp2.mlp.layer0.bn.bn.num_batches_tracked", "backbone_net4.fp2.mlp.layer1.conv.weight", "backbone_net4.fp2.mlp.layer1.bn.bn.weight", "backbone_net4.fp2.mlp.layer1.bn.bn.bias", "backbone_net4.fp2.mlp.layer1.bn.bn.running_mean", "backbone_net4.fp2.mlp.layer1.bn.bn.running_var", "backbone_net4.fp2.mlp.layer1.bn.bn.num_batches_tracked", "conv_agg1.weight", "conv_agg1.bias", "bn_agg1.weight", "bn_agg1.bias", "bn_agg1.running_mean", "bn_agg1.running_var", "bn_agg1.num_batches_tracked", "conv_agg2.weight", "conv_agg2.bias", "bn_agg2.weight", "bn_agg2.bias", "bn_agg2.running_mean", "bn_agg2.running_var", "bn_agg2.num_batches_tracked", "conv_flag_z1.weight", "conv_flag_z1.bias", "bn_flag_z1.weight", "bn_flag_z1.bias", "bn_flag_z1.running_mean", "bn_flag_z1.running_var", "bn_flag_z1.num_batches_tracked", "conv_flag_z2.weight", "conv_flag_z2.bias", "conv_flag_xy1.weight", "conv_flag_xy1.bias", "bn_flag_xy1.weight", "bn_flag_xy1.bias", "bn_flag_xy1.running_mean", "bn_flag_xy1.running_var", "bn_flag_xy1.num_batches_tracked", "conv_flag_xy2.weight", "conv_flag_xy2.bias", "conv_flag_line1.weight", "conv_flag_line1.bias", "bn_flag_line1.weight", "bn_flag_line1.bias", "bn_flag_line1.running_mean", "bn_flag_line1.running_var", "bn_flag_line1.num_batches_tracked", "conv_flag_line2.weight", "conv_flag_line2.bias", "vgen.conv1.weight", "vgen.conv1.bias", "vgen.conv2.weight", "vgen.conv2.bias", "vgen.conv3.weight", "vgen.conv3.bias", "vgen.bn1.weight", "vgen.bn1.bias", "vgen.bn1.running_mean", "vgen.bn1.running_var", "vgen.bn1.num_batches_tracked", "vgen.bn2.weight", "vgen.bn2.bias", "vgen.bn2.running_mean", "vgen.bn2.running_var", "vgen.bn2.num_batches_tracked", "vgen_z.conv1.weight", "vgen_z.conv1.bias", "vgen_z.conv2.weight", "vgen_z.conv2.bias", "vgen_z.conv3.weight", "vgen_z.conv3.bias", "vgen_z.bn1.weight", "vgen_z.bn1.bias", "vgen_z.bn1.running_mean", "vgen_z.bn1.running_var", "vgen_z.bn1.num_batches_tracked", "vgen_z.bn2.weight", "vgen_z.bn2.bias", "vgen_z.bn2.running_mean", "vgen_z.bn2.running_var", "vgen_z.bn2.num_batches_tracked", "vgen_xy.conv1.weight", "vgen_xy.conv1.bias", "vgen_xy.conv2.weight", "vgen_xy.conv2.bias", "vgen_xy.conv3.weight", "vgen_xy.conv3.bias", "vgen_xy.bn1.weight", "vgen_xy.bn1.bias", "vgen_xy.bn1.running_mean", "vgen_xy.bn1.running_var", "vgen_xy.bn1.num_batches_tracked", "vgen_xy.bn2.weight", "vgen_xy.bn2.bias", "vgen_xy.bn2.running_mean", "vgen_xy.bn2.running_var", "vgen_xy.bn2.num_batches_tracked", "vgen_line.conv1.weight", "vgen_line.conv1.bias", "vgen_line.conv2.weight", "vgen_line.conv2.bias", "vgen_line.conv3.weight", "vgen_line.conv3.bias", "vgen_line.bn1.weight", "vgen_line.bn1.bias", "vgen_line.bn1.running_mean", "vgen_line.bn1.running_var", "vgen_line.bn1.num_batches_tracked", "vgen_line.bn2.weight", "vgen_line.bn2.bias", "vgen_line.bn2.running_mean", "vgen_line.bn2.running_var", "vgen_line.bn2.num_batches_tracked", "pnet_z.vote_aggregation.mlp_module.layer0.conv.weight", "pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.weight", "pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.bias", "pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "pnet_z.vote_aggregation.mlp_module.layer0.bn.bn.num_batches_tracked", "pnet_z.vote_aggregation.mlp_module.layer1.conv.weight", "pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.weight", "pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.bias", "pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "pnet_z.vote_aggregation.mlp_module.layer1.bn.bn.num_batches_tracked", "pnet_z.vote_aggregation.mlp_module.layer2.conv.weight", "pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.weight", "pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.bias", "pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "pnet_z.vote_aggregation.mlp_module.layer2.bn.bn.num_batches_tracked", "pnet_z.conv1.weight", "pnet_z.conv1.bias", "pnet_z.conv2.weight", "pnet_z.conv2.bias", "pnet_z.conv3.weight", "pnet_z.conv3.bias", "pnet_z.bn1.weight", "pnet_z.bn1.bias", "pnet_z.bn1.running_mean", "pnet_z.bn1.running_var", "pnet_z.bn1.num_batches_tracked", "pnet_z.bn2.weight", "pnet_z.bn2.bias", "pnet_z.bn2.running_mean", "pnet_z.bn2.running_var", "pnet_z.bn2.num_batches_tracked", "pnet_xy.vote_aggregation.mlp_module.layer0.conv.weight", "pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.weight", "pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.bias", "pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "pnet_xy.vote_aggregation.mlp_module.layer0.bn.bn.num_batches_tracked", "pnet_xy.vote_aggregation.mlp_module.layer1.conv.weight", "pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.weight", "pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.bias", "pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "pnet_xy.vote_aggregation.mlp_module.layer1.bn.bn.num_batches_tracked", "pnet_xy.vote_aggregation.mlp_module.layer2.conv.weight", "pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.weight", "pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.bias", "pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "pnet_xy.vote_aggregation.mlp_module.layer2.bn.bn.num_batches_tracked", "pnet_xy.conv1.weight", "pnet_xy.conv1.bias", "pnet_xy.conv2.weight", "pnet_xy.conv2.bias", "pnet_xy.conv3.weight", "pnet_xy.conv3.bias", "pnet_xy.bn1.weight", "pnet_xy.bn1.bias", "pnet_xy.bn1.running_mean", "pnet_xy.bn1.running_var", "pnet_xy.bn1.num_batches_tracked", "pnet_xy.bn2.weight", "pnet_xy.bn2.bias", "pnet_xy.bn2.running_mean", "pnet_xy.bn2.running_var", "pnet_xy.bn2.num_batches_tracked", "pnet_line.vote_aggregation.mlp_module.layer0.conv.weight", "pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.weight", "pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.bias", "pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "pnet_line.vote_aggregation.mlp_module.layer0.bn.bn.num_batches_tracked", "pnet_line.vote_aggregation.mlp_module.layer1.conv.weight", "pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.weight", "pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.bias", "pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "pnet_line.vote_aggregation.mlp_module.layer1.bn.bn.num_batches_tracked", "pnet_line.vote_aggregation.mlp_module.layer2.conv.weight", "pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.weight", "pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.bias", "pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "pnet_line.vote_aggregation.mlp_module.layer2.bn.bn.num_batches_tracked", "pnet_line.conv1.weight", "pnet_line.conv1.bias", "pnet_line.conv2.weight", "pnet_line.conv2.bias", "pnet_line.conv3.weight", "pnet_line.conv3.bias", "pnet_line.bn1.weight", "pnet_line.bn1.bias", "pnet_line.bn1.running_mean", "pnet_line.bn1.running_var", "pnet_line.bn1.num_batches_tracked", "pnet_line.bn2.weight", "pnet_line.bn2.bias", "pnet_line.bn2.running_mean", "pnet_line.bn2.running_var", "pnet_line.bn2.num_batches_tracked", "pnet_final.vote_aggregation.mlp_module.layer0.conv.weight", "pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.weight", "pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.bias", "pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.running_mean", "pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.running_var", "pnet_final.vote_aggregation.mlp_module.layer0.bn.bn.num_batches_tracked", "pnet_final.vote_aggregation.mlp_module.layer1.conv.weight", "pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.weight", "pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.bias", "pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.running_mean", "pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.running_var", "pnet_final.vote_aggregation.mlp_module.layer1.bn.bn.num_batches_tracked", "pnet_final.vote_aggregation.mlp_module.layer2.conv.weight", "pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.weight", "pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.bias", "pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.running_mean", "pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.running_var", "pnet_final.vote_aggregation.mlp_module.layer2.bn.bn.num_batches_tracked", "pnet_final.match_surface_center.mlp_module.layer0.conv.weight", "pnet_final.match_surface_center.mlp_module.layer0.bn.bn.weight", "pnet_final.match_surface_center.mlp_module.layer0.bn.bn.bias", "pnet_final.match_surface_center.mlp_module.layer0.bn.bn.running_mean", "pnet_final.match_surface_center.mlp_module.layer0.bn.bn.running_var", "pnet_final.match_surface_center.mlp_module.layer0.bn.bn.num_batches_tracked", "pnet_final.match_surface_center.mlp_module.layer1.conv.weight", "pnet_final.match_surface_center.mlp_module.layer1.bn.bn.weight", "pnet_final.match_surface_center.mlp_module.layer1.bn.bn.bias", "pnet_final.match_surface_center.mlp_module.layer1.bn.bn.running_mean", "pnet_final.match_surface_center.mlp_module.layer1.bn.bn.running_var", "pnet_final.match_surface_center.mlp_module.layer1.bn.bn.num_batches_tracked", "pnet_final.match_surface_center.mlp_module.layer2.conv.weight", "pnet_final.match_surface_center.mlp_module.layer2.bn.bn.weight", "pnet_final.match_surface_center.mlp_module.layer2.bn.bn.bias", "pnet_final.match_surface_center.mlp_module.layer2.bn.bn.running_mean", "pnet_final.match_surface_center.mlp_module.layer2.bn.bn.running_var", "pnet_final.match_surface_center.mlp_module.layer2.bn.bn.num_batches_tracked", "pnet_final.match_line_center.mlp_module.layer0.conv.weight", "pnet_final.match_line_center.mlp_module.layer0.bn.bn.weight", "pnet_final.match_line_center.mlp_module.layer0.bn.bn.bias", "pnet_final.match_line_center.mlp_module.layer0.bn.bn.running_mean", "pnet_final.match_line_center.mlp_module.layer0.bn.bn.running_var", "pnet_final.match_line_center.mlp_module.layer0.bn.bn.num_batches_tracked", "pnet_final.match_line_center.mlp_module.layer1.conv.weight", "pnet_final.match_line_center.mlp_module.layer1.bn.bn.weight", "pnet_final.match_line_center.mlp_module.layer1.bn.bn.bias", "pnet_final.match_line_center.mlp_module.layer1.bn.bn.running_mean", "pnet_final.match_line_center.mlp_module.layer1.bn.bn.running_var", "pnet_final.match_line_center.mlp_module.layer1.bn.bn.num_batches_tracked", "pnet_final.match_line_center.mlp_module.layer2.conv.weight", "pnet_final.match_line_center.mlp_module.layer2.bn.bn.weight", "pnet_final.match_line_center.mlp_module.layer2.bn.bn.bias", "pnet_final.match_line_center.mlp_module.layer2.bn.bn.running_mean", "pnet_final.match_line_center.mlp_module.layer2.bn.bn.running_var", "pnet_final.match_line_center.mlp_module.layer2.bn.bn.num_batches_tracked", "pnet_final.conv1.weight", "pnet_final.conv1.bias", "pnet_final.conv2.weight", "pnet_final.conv2.bias", "pnet_final.conv3.weight", "pnet_final.conv3.bias", "pnet_final.bn1.weight", "pnet_final.bn1.bias", "pnet_final.bn1.running_mean", "pnet_final.bn1.running_var", "pnet_final.bn1.num_batches_tracked", "pnet_final.bn2.weight", "pnet_final.bn2.bias", "pnet_final.bn2.running_mean", "pnet_final.bn2.running_var", "pnet_final.bn2.num_batches_tracked", "pnet_final.conv_match1.weight", "pnet_final.conv_match1.bias", "pnet_final.conv_match2.weight", "pnet_final.conv_match2.bias", "pnet_final.bn_match1.weight", "pnet_final.bn_match1.bias", "pnet_final.bn_match1.running_mean", "pnet_final.bn_match1.running_var", "pnet_final.bn_match1.num_batches_tracked", "pnet_final.conv_match_sem1.weight", "pnet_final.conv_match_sem1.bias", "pnet_final.conv_match_sem2.weight", "pnet_final.conv_match_sem2.bias", "pnet_final.bn_match_sem1.weight", "pnet_final.bn_match_sem1.bias", "pnet_final.bn_match_sem1.running_mean", "pnet_final.bn_match_sem1.running_var", "pnet_final.bn_match_sem1.num_batches_tracked", "pnet_final.conv_surface1.weight", "pnet_final.conv_surface1.bias", "pnet_final.conv_surface2.weight", "pnet_final.conv_surface2.bias", "pnet_final.bn_surface1.weight", "pnet_final.bn_surface1.bias", "pnet_final.bn_surface1.running_mean", "pnet_final.bn_surface1.running_var", "pnet_final.bn_surface1.num_batches_tracked", "pnet_final.bn_surface2.weight", "pnet_final.bn_surface2.bias", "pnet_final.bn_surface2.running_mean", "pnet_final.bn_surface2.running_var", "pnet_final.bn_surface2.num_batches_tracked", "pnet_final.conv_line1.weight", "pnet_final.conv_line1.bias", "pnet_final.conv_line2.weight", "pnet_final.conv_line2.bias", "pnet_final.bn_line1.weight", "pnet_final.bn_line1.bias", "pnet_final.bn_line1.running_mean", "pnet_final.bn_line1.running_var", "pnet_final.bn_line1.num_batches_tracked", "pnet_final.bn_line2.weight", "pnet_final.bn_line2.bias", "pnet_final.bn_line2.running_mean", "pnet_final.bn_line2.running_var", "pnet_final.bn_line2.num_batches_tracked", "pnet_final.conv_refine1.weight", "pnet_final.conv_refine1.bias", "pnet_final.conv_refine2.weight", "pnet_final.conv_refine2.bias", "pnet_final.conv_refine3.weight", "pnet_final.conv_refine3.bias", "pnet_final.conv_refine4.weight", "pnet_final.conv_refine4.bias", "pnet_final.bn_refine1.weight", "pnet_final.bn_refine1.bias", "pnet_final.bn_refine1.running_mean", "pnet_final.bn_refine1.running_var", "pnet_final.bn_refine1.num_batches_tracked", "pnet_final.bn_refine2.weight", "pnet_final.bn_refine2.bias", "pnet_final.bn_refine2.running_mean", "pnet_final.bn_refine2.running_var", "pnet_final.bn_refine2.num_batches_tracked", "pnet_final.bn_refine3.weight", "pnet_final.bn_refine3.bias", "pnet_final.bn_refine3.running_mean", "pnet_final.bn_refine3.running_var", "pnet_final.bn_refine3.num_bat

i'm confused about that

@GitBoSun @zaiweizhang
Hi,
In file sunrgbd/sunrgbd_detection_dataset_hd, it is mentioned in comments that, sizes l,w,h are half the box dimensions and angle is measured from +x to -y.
While, in file utils/show_results_sunrgbd.py, the box sizes l,w,h seems to be taken as equal to the predicted/gt box dimensions.
Please comment on exactly what sizes l,w,h, and the heading angle represent in the input labels and in the predicted labels by the model.

Hi Zaiwei,

Thanks for sharing your work. I have a few questions on your implementation details, which are not consistent with the paper.

1. In Section 3.2, you mentioned that 0.2m is used to select positive points lying close to BB face or BB edge. However, 0.1 is used for SUN RGB-D dataset in the code:
   

   H3DNet/sunrgbd/sunrgbd_detection_dataset_hd.py

   Line 40 in e89d092

   DIST_THRESH = 0.1#0.2

   
   How do you determine the value of this threshold? And How will it affect the performance?

2. Also in Section 3.2, you mentioned that

The predicted attributes include a flag that indicates whether a point is close to a BB face or not and if so, an offset vector between that point and its corresponding BB face center.

However, additional surface size is predicted in the code:

May I know how it (i.e., adding size loss for surfaces) improves the performance, especially on SUN RGB-D?

3. When generating ground-truths for point boundary offsets, it is possible that one point can be close to two BB faces (e.g., right and front faces). However, the code seems overwriting the offset even if one point might have already been assigned to another face (e.g., L569-575 in sunrgbd_detection_dataset_hd.py). Am I right about this? If it is true, how much will it affect the performance?

Looking forward to your reply. Thanks.

Hi,

Have a good day to you. I would like to seek your help for find pretrained model of H3DNet to testing purpose only.
Thank you very much.

**** EPOCH 000 ****
Current learning rate: 0.010000
Current BN decay momentum: 0.500000
2021-03-17 08:52:17.518997
Traceback (most recent call last):
File "train.py", line 379, in
train(start_epoch)
File "train.py", line 358, in train
train_one_epoch()
File "train.py", line 263, in train_one_epoch
loss.backward()
File "/usr/local/lib/python3.6/dist-packages/torch/tensor.py", line 195, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph)
File "/usr/local/lib/python3.6/dist-packages/torch/autograd/init.py", line 99, in backward
allow_unreachable=True) # allow_unreachable flag
RuntimeError: Expected isFloatingType(grads[i].type().scalarType()) to be true, but got false. (Could this error message be improved? If so, please report an enhancement request to PyTorch.)

I tried to reproduce as your instruction but the bug happens, do you have any idea why it did ? Thank you very much.

Hi @zaiweizhang and @GitBoSun,
Thank you for this repo.
I am trying to see the visualizations, but when I ran show_results_scannet.py, I could only get the bounding boxes, but how do I get the scene and bounding boxes on top of it. can you please help me to get this issue solved for me.
Thank you

Hi

I am getting following errors when starts training. Do you know what might be the reason?!

Ubuntu16
pytorch 1.1
tensorflow-gpu=1.14
cuda=10
cudnn=7.4
GPU=RTX2080ti

Thank You

**** EPOCH 000 ****
Current learning rate: 0.001000
Current BN decay momentum: 0.500000
2020-08-02 19:38:48.510935
THCudaCheck FAIL file=/pytorch/aten/src/THC/THCGeneral.cpp line=383 error=11 : invalid argument
Traceback (most recent call last):
  File "train.py", line 382, in <module>
    train(start_epoch)
  File "train.py", line 361, in train
    train_one_epoch()
  File "train.py", line 257, in train_one_epoch
    end_points = net(inputs, end_points)            
  File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/sahar/Mohammad_ws/H3DNet/models/hdnet.py", line 185, in forward
    end_points = self.pnet_final(proposal_xyz, proposal_features, center_z, feature_z, center_xy, feature_xy, center_line, feature_line, end_points)
  File "/usr/local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/sahar/Mohammad_ws/H3DNet/models/proposal_module_refine.py", line 276, in forward
    obj_surface_center, obj_line_center = get_surface_line_points_batch_pytorch(obj_size, pred_heading, obj_center)
  File "/home/sahar/Mohammad_ws/H3DNet/utils/box_util.py", line 353, in get_surface_line_points_batch_pytorch
    surface_3d = torch.matmul(surface_3d.unsqueeze(-2), surface_rot.transpose(3,2)).squeeze(-2)
RuntimeError: cublas runtime error : the GPU program failed to execute at /pytorch/aten/src/THC/THCBlas.cu:450

Hi zaiweizhang,

When I am trying to make visualization on the result of a checkpoint, I found that the PRED_PATH in show_results_sunrgbd.py is confused. After I execute the eval.py, there is no record of predictions.

Could you please give me some advices to set the PRED_PATH or construct the prediction files?

Thanks,
Ke