nv-tlabs / get3d Goto Github PK

View Code? Open in Web Editor NEW

4.1K 4.1K 369.0 175.36 MB

License: Other

Python 80.58% Dockerfile 0.23% Shell 0.16% C++ 4.95% Cuda 14.08%

get3d's People

Contributors

Stargazers

Watchers

Forkers

babajideowoyele fastflair furmanlukasz recreatemyself elementalsparke alexjoonn doutdex markjgx cl0ckt0wer zakfong zivzone archiyujie likeucode sea-conch milky916 clickbox bobthehands jaedukseo perfyperfect lucascviveiros lavosbit elevate-labs onthegocode ocipower bruinxiong xiaohunqupo bonbinker fabian2112 luminova-japan 14king richardsonjf radychs tli347 nanyu1226 ricklentz kakzaki cuulee wuzerun-888 imraj codeaudit rsh4d0w kolourr 0xsmw hpp340 nongkris brechtcorbeel n1ckfg genshare snapbuy super-rain suhail b1sounours voletiv armandocastro2 smeshing olegjakushkin sapium59 lucianna thibaultgroueix goldinium gcabc123 peterzs andrewfowl orangesodahub toriningen popanku ashwinrajendraprasad webaverse atlasgooo2 joymagine lalalune mindrages tontontremblay easychart bapek jags111 cny689 lupin4 ericktornero anoop-qasolve purplesparkle hax0rgurl cabelo mu-l jonyhuang leanderdulac liqizha acondess leoxing1996 odweibo walklook ausjorg mirageml adambear freja71122 matthewsanders philmaas maskinoshita neverix developerjigr

get3d's Issues

Persue higher resolution for mesh

I am happy to see pre-trained model released! Thank you!

Now I have infered a few times with checkpoint of car category. As far as I observe, the triangular meshes has nearly the same resolution of about 5 cm, or 5% of model total length. Is there a way to make more detailed meshes?

Also, I guess your mesh reconstruction is based on SDF-driven DMTet algorithm. Do you expect bad behavior due to inaccuracy of SDF when applyibg finer resolution?

p.s. My inference took 20 mins to produce 100 textured meshes, so computation complexity do not quite impact me.

Final step of environment setup / WGET

First of all, congratulations to the team for such groundbreaking work! I'm not a programmer but 20+ years in CGI have forced me to try and understand code. I'm on Win10, using Anaconda, my environment is pretty much configured, except that on the very last step according to the readme, I get an error when running this wget command:

wget https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/versions/1/files/metrics/inception-2015-12-05.pkl

The error states the identity of the server could not be confirmed and suggests using a flag to bypass authentication but ngc rejects such a connection. I've gone through the NGC documentation and in the NGC library we can generate wget or curl commands to pull stuff but they are identical to what the readme suggests. I've created an NGC account and gone through the setup, which generates a config file, but haven't been able to use that when calling wget. I know this is not even related to the project but for other up-and-coming W10 users this might be a point where they get stuck. Any pointers?

Thanks in advance,
L

I have a bug that is fatal error: EGL/egl.h: No such file or directory.

I met this bug when I try training. Here are the details.

[3/4] c++ -MMD -MF glutil.o.d -DTORCH_EXTENSION_NAME=nvdiffrast_plugin_gl -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="_gcc" -DPYBIND11_STDLIB="_libstdcpp" -DPYBIND11_BUILD_ABI="_cxxabi1011" -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include/torch/csrc/api/include -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include/TH -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /SSD_DISK/users/anaconda3/envs/get3d1/include/python3.8 -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++14 -DNVDR_TORCH -c /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/nvdiffrast/common/glutil.cpp -o glutil.o
FAILED: glutil.o
c++ -MMD -MF glutil.o.d -DTORCH_EXTENSION_NAME=nvdiffrast_plugin_gl -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="_gcc" -DPYBIND11_STDLIB="_libstdcpp" -DPYBIND11_BUILD_ABI="_cxxabi1011" -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include/torch/csrc/api/include -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include/TH -isystem /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /SSD_DISK/users/anaconda3/envs/get3d1/include/python3.8 -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++14 -DNVDR_TORCH -c /SSD_DISK/users/qianjiachen/anaconda3/envs/get3d1/lib/python3.8/site-packages/nvdiffrast/common/glutil.cpp -o glutil.o
In file included from /SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/nvdiffrast/common/glutil.cpp:14:
/SSD_DISK/users/anaconda3/envs/get3d1/lib/python3.8/site-packages/nvdiffrast/common/glutil.h:36:10: fatal error: EGL/egl.h: No such file or directory
36 | #include <EGL/egl.h>
| ^~~~~~~~~~~
compilation terminated.

Random Seed

Hi @SteveJunGao how can I use --seed to change the order of output from inference?

Remove hardcoded values for datasets and add args for dataset input

Right now there are many references in the code to specific datasets, as well as a data_camera_mode option that takes a dataset name which gets handled with some complex if statements at various points in the training code.

Raising the issue here to remove these specific requirements and generalize modes ("object", "human", etc). Maybe we pass a configuration file in so that future researchers can add their own custom parameters.

(by the way, I'm working on a PR for this!)

Release of trained models

Hi,

I'm wondering if there are plans to release trained models so we can play with the trained models for inference only.

Thanks.

no lfd feature calculated using compute_lfd_feat_multiprocess.py

Hi there, I've been testing the evaluation metrics using 2 example dataset which contains 2 ground truth obj, and 20 generated obj.

It seems that

Step 4: We first generat the Light Field feature for each object by running

python compute_lfd_feat_multiprocess.py --gen_path PATH_TO_THE_MODEL_PREDICTION --save_path PATH_FOR_LFD_OUTPUT_FOR_PRED

only create null files like mesh_q4_v_1.8.art, mesh_q8_v1.8.art, etc.

Then when I try to run the compute_lfd scripts: using python compute_lfd.py --split_path PATH_TO_TEST_SPLIT --dataset_path PATH_FOR_LFD_OUTPUT_FOR_GT --gen_path PATH_FOR_LFD_OUTPUT_FOR_PRED --save_name results/our/lfd.pkl
I got Segmentation fault (core dumped) when loading the data.

GET3D/evaluation_scripts/compute_lfd.py

Line 25 in 6fda9cd

result = load_data.run(os.path.join(f, 'mesh'))

Is the data supposed to be null? or it should be the calculated lfd feature? I've roughly checked the align_mesh() function in lfd_me.py but it seems that I couldn't find a line of code that actually write lfd feature to the created null files: mesh_q4_v_1.8.art, mesh_q8_v1.8.art, etc.

Appreciate your help!

"Connection refused" error when training

Hi,
I want to train this network on my own dataset (which I have processed using the provided rendering scripts) however I get an "ConnectionRefusedError: [Errno 111] Connection refused" error and an "NotImplementedError" raised in the discriminator_architecture script when I run the training. What might be the issue here?

Support for ShapeNet V2

Question--

is there any reason, other than normalizing folder structure, that ShapeNetCore V2 wouldn't work, or reason you chose ShapeNet V1 for the release?
Would you accept a PR for support for ShapeNet V2?

Not Found Camera Root

When training with the following command:

python train_3d.py --outdir=results/ --data=render_shapenet_data/dataset/img/02958343/ --camera_path=render_shapenet_data/dataset/camera/02958343/ --gpus=8 --batch=32 --gamma=40 --data_camera_mode shapenet_car  --dmtet_scale 1.0  --use_shapenet_split 1  --one_3d_generator 1  --fp32 0

I keep getting this output:

20%|##        | 57/279 [00:51<02:59,  1.24it/s]==> not found camera root
==> not found camera root
==> not found camera root

Do I need to fix something?

How to use labels in training + how to use CLIP guidance in inferrence

In the paper this is mentioned, however it is not clear how to get this working.

Run inference on a specific image

Right now, inference is run on the eval images.

It'd be awfully handy to be able to assess arbitrary images by passing one to the CLI!

shapenet_chair inference

I see shapenetcorev1 has chair images. When we run inference on shapenet_chair, I see log messages that ./tmp has 1234 images but I cannot see those. Also val.txt in 3dgan_data_split folder has 573 ids. When we run inference, what is the input image? The fakes_000000_00.png file has 25 generated images. mesh_pred folder has 10 obj files. Can you please explain what is the correlation between these? What was the input image to the inference, in other words, what was the 2D image for which the 3D was generated?

Could you provide the pretrained weights for animal categories?

Thanks a lot!

Multiple classes in same checkpoint

I am hoping to train a general model on 3D understanding beyond just a single class or range of classes.

I am doing this by resuming training on a checkpoint while swapping out datasets, however I am unsure of the limitations here.

Will training on a wider range of data make the model worse at individual tasks? Or is there otherwise any reason not to do this?

Volume Sub-division

Where is subdivision used in the code? The function batch_subdivide_volume doesn't seem to be used.

Text-Guided 3D Synthesis

Hi @SteveJunGao thank you for all your help!
When you release the pretrained model (Issue #16) will you also be releasing the weights for the fine-tuned model on CLIP?
Excited to continue working on this!

Seed

2d images to 3d model

Hello everyone, I did not find in the article a description of how to make 3d models from 2d pictures, just a little about the verbal description of the style. Logically, of course, it can be assumed that instead of a sample from the distribution, it is possible to submit a picture to the MLP input, but due to the closeness of the architecture, I could not do this experience. Perhaps someone has already tried to conduct a similar experience? Could you tell us how to do this on this project ?

Material Generation Supported?

Hi, I am interested in the material generation ability described in the paper. However, I can't find relevant code relating to either rendering images using material properties or generating real images from Turbosquid models (with consistent materials) using Blender.
Did I miss something in the repo? Or the material generation code is not available yet?

SDF regularization loss

GET3D/uni_rep/rep_3d/dmtet.py

Line 371 in 18c35f1

def sdf_reg_loss(sdf, all_edges):

The paper mentions using this loss. In the code i see this function is not called anywhere. Where is this supposed to fit in?

Thanks!

Using Blender Eevee and HDRI lighting for rendering ShapeNet

We've modified the input scripts and are starting on our journey of training the models:
https://github.com/webaverse/GET3D/blob/master/render_shapenet_data/render_shapenet.py

We got much faster and more game-like generation results by switching to Eevee renderer and going to HDRI. Eevee renders 24-frame multiview in ~3-4 seconds per model, and it can crank through an entire average-sized shape category set in about an hour or two.

Also, because it's a realtime-style rasterizing renderer, it's pixel perfect and not subject to noise. To compensate for a lack of dimension we've lit the scene with a studio HDRI and applied an ambient occlusion post processing effect.

So, questions for the team:

do you see any issues with this, is there a requirement for the realistic renderer?
would you accept a pull request for this? we will need to modify our code since we are also doing some transformation to normalize the ShapeNet V2 dataset, but I could make a PR to switch between eevee and cycles.

About the pretrain model

Maybe it's my fault that I didn't find the download address of the pre training model. This training scale is too large for me. I want to directly infer to see the effect of the network.I want to ask if there is a trained model that can be directly used for inference.

Issue on training

Hi There,

I followed the training guide in README，and I met this problem: TypeError: unsupported operand type(s) for /: 'Nonetype' and 'int'
Do you have any ideas how to fix this problem?

My training script is like this:

Google Colab for Artists/Game Designers

Hi There,

Have had some good success with a dreamfields collab. Please see here: https://www.instagram.com/p/CjURPkWrMvh/ . Was actually able to import it into unreal; please see here: https://www.instagram.com/p/CjUZS2zuvRS/ . Would it be possible to get a colab of this work so that we, artists/game designers, can begin experimenting with it? Thank you!

Hi Are you going to add zero-shot single image to 3D model , or text to 3d model?

Are you going to add zero-shot single image to 3D model , or text to 3d model?

Can this model fit Google Drive?

wget https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/versions/1/files/metrics/inception-2015-12-05.pkl

Please send me the download link below, thank you!

Is this a mistake when parse the "cammv" from tuple"?

network_get3d:466
network_get3d:470

Support for DDP with multi nodes

Hi there,

I found the code can only run with multi gpus in one node, if there are supports for multi gpus in multi nodes?

@SteveJunGao

How to set gamma to generate objects of the bowl category?

Thanks for sharing very wonderful work to generate synthetic data.

Now I recognize that I should set gamma according to the complexity of dataset from the issue No.15(#15).

But could you recommend roughly "gamma" value for the "bowl" category data of ShapeNet v1 dataset?

Sample images of "bowl" category look like below.

Thanks in advance.

Setting up PyTorch plugin "upfirdn2d_plugin"... Failed!

Can you help me to train model.
When I try train model with car dataset, I met this error

Bug in inference with lower resolution

Hi,
Thanks for you awesome work!
I am trying to use your code to train the model with image resolution = 512x512. However, it seems this line of code
always assumes the resolution with 1024x1024, and thus I cannot do inference for my trained model. Maybe it should be fixed~

Deformation prediction

Hi,
Thanks for your great work! I have a question about the deformation prediction:
I noticed that in your previous work (i.e. DEFTet and DMTet), you used GCN to predict the deformation. However, in GET3D, you use a simple MLP as your network architecture. Could you please give some explainations for this modification?
Thank you.

What is the minimum number of images required during inference time?

About the results of the pretrained model

I'm so sorry to bother you again. But I believe that you are so enthusiastic that you could solve my confusion.
I infer GET3D with the pretrained weight named shapenet_xx.pt and others on colab. But the results is not as good as your presentation, especially the 3D mesh of motorbikes. The stickiness of details is high. Even so, it's a good work. The steps i took are as your readme file. Is there anything i missed or ignored? or maybe i should train the model by myself?

Where is the generated model? Is it fbx or obj?

Hi
Where is the generated model? Is it fbx or obj? If not, how can I use the generated 3d mesh model?

If I train the model myself, what do I need? A lot of png, or a lot of mp4 videos?

Do you have a specific use tutorial?

any support for Mac M1?

congrats on an amazing achievement! i'm on a MacStudio M1 Max with 64 GB of shared RAM so i can run it on 16GB. can this be run as is or has anyone actually tried running it on a Mac M1/M2? if not is there any plan for a Google Colab version? also wondering if there's a forum to get info from others trying this out?

How to set gamma and dmtet_scale for different data?

Thanks for sharing your awesome work!

Since you gave different settings for different classes, I am wondering if you have any insight to tune the 'gamma' and 'dmtet_scale' in your training script? Currently we are trying to implement your framework to other datasets, your advice would be very helpful.

Thank you.

Regarding dataset

This is a very good work. Congratulations! Would you please provide relevant experimental datasets, such as the House dataset (563 shapes) collected from Turbosquid? Thanks in advance.

How to feed in image as inference input?

My understanding is that a single image, in form of output of inception-v3 encoder, can serve as input for inference input. I also noticed the pkl file of the pre-trained inception model you offered as a part of necessary files. I am newbie. I simply don't know how to utilize them. Is there some way to feed the image (or its latent code) into inference script?
Thank you!

Any Colab Setup

Hi someone
i try make one colab

https://colab.research.google.com/gist/kilik128/9915b90fd8f405f96da36422ecf30f43/get3d.ipynb

look crash at this time
any idea as welcome
thank's

Issue on rendering Shapenet Dataset

When running:
python render_all.py --save_folder dataset/ --dataset_folder ShapeNetCore.v1/ --blender_root <path-to-blender>

Getting the following error:

ALSA lib confmisc.c:767:(parse_card) cannot find card '0'
ALSA lib conf.c:4732:(_snd_config_evaluate) function snd_func_card_driver returned error: No such file or directory
ALSA lib confmisc.c:392:(snd_func_concat) error evaluating strings
ALSA lib conf.c:4732:(_snd_config_evaluate) function snd_func_concat returned error: No such file or directory
ALSA lib confmisc.c:1246:(snd_func_refer) error evaluating name
ALSA lib conf.c:4732:(_snd_config_evaluate) function snd_func_refer returned error: No such file or directory
ALSA lib conf.c:5220:(snd_config_expand) Evaluate error: No such file or directory
ALSA lib pcm.c:2642:(snd_pcm_open_noupdate) Unknown PCM default

Inference for other Shapenet Datasets

Hi,
I trained on the shapenet_rocket dataset using this PR: #23
I am trying to do inference on the model I trained however I am getting this error:

==> resume from pretrained path checkpoints/shapenet_rocket/00007-stylegan2-04099429-gpus8-batch32-gamma80/network-snapshot-001843.pt
Traceback (most recent call last):
  File "train_3d.py", line 319, in <module>
    main()  # pylint: disable=no-value-for-parameter
  File "/opt/conda/lib/python3.8/site-packages/click/core.py", line 829, in __call__
    return self.main(*args, **kwargs)
  File "/opt/conda/lib/python3.8/site-packages/click/core.py", line 782, in main
    rv = self.invoke(ctx)
  File "/opt/conda/lib/python3.8/site-packages/click/core.py", line 1066, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/opt/conda/lib/python3.8/site-packages/click/core.py", line 610, in invoke
    return callback(*args, **kwargs)
  File "train_3d.py", line 313, in main
    launch_training(c=c, desc=desc, outdir=opts.outdir, dry_run=opts.dry_run)
  File "train_3d.py", line 103, in launch_training
    subprocess_fn(rank=0, c=c, temp_dir=temp_dir)
  File "train_3d.py", line 46, in subprocess_fn
    inference_3d.inference(rank=rank, **c)
  File "/workspace/GET3D/training/inference_3d.py", line 81, in inference
    G.load_state_dict(model_state_dict['G'], strict=True)
  File "/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1406, in load_state_dict
    raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for GeneratorDMTETMesh:
        Missing key(s) in state_dict: "synthesis.generator.tri_plane_synthesis.b4.const", "synthesis.generator.tri_plane_synthesis.b4.resample_filter", "synthesis.generator.tri_plane_synthesis.b4.conv1.weight", "synthesis.generator.tri_plane_synthesis.b4.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b4.conv1.bias", "synthesis.generator.tri_plane_synthesis.b4.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b4.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b4.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b4.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b4.totex.weight", "synthesis.generator.tri_plane_synthesis.b4.totex.bias", "synthesis.generator.tri_plane_synthesis.b4.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b4.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b4.togeo.weight", "synthesis.generator.tri_plane_synthesis.b4.togeo.bias", "synthesis.generator.tri_plane_synthesis.b4.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b4.togeo.affine.bias", "synthesis.generator.tri_plane_synthesis.b8.resample_filter", "synthesis.generator.tri_plane_synthesis.b8.conv0.weight", "synthesis.generator.tri_plane_synthesis.b8.conv0.noise_strength", "synthesis.generator.tri_plane_synthesis.b8.conv0.bias", "synthesis.generator.tri_plane_synthesis.b8.conv0.resample_filter", "synthesis.generator.tri_plane_synthesis.b8.conv0.noise_const", "synthesis.generator.tri_plane_synthesis.b8.conv0.affine.weight", "synthesis.generator.tri_plane_synthesis.b8.conv0.affine.bias", "synthesis.generator.tri_plane_synthesis.b8.conv1.weight", "synthesis.generator.tri_plane_synthesis.b8.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b8.conv1.bias", "synthesis.generator.tri_plane_synthesis.b8.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b8.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b8.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b8.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b8.totex.weight", "synthesis.generator.tri_plane_synthesis.b8.totex.bias", "synthesis.generator.tri_plane_synthesis.b8.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b8.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b8.togeo.weight", "synthesis.generator.tri_plane_synthesis.b8.togeo.bias", "synthesis.generator.tri_plane_synthesis.b8.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b8.togeo.affine.bias", "synthesis.generator.tri_plane_synthesis.b16.resample_filter", "synthesis.generator.tri_plane_synthesis.b16.conv0.weight", "synthesis.generator.tri_plane_synthesis.b16.conv0.noise_strength", "synthesis.generator.tri_plane_synthesis.b16.conv0.bias", "synthesis.generator.tri_plane_synthesis.b16.conv0.resample_filter", "synthesis.generator.tri_plane_synthesis.b16.conv0.noise_const", "synthesis.generator.tri_plane_synthesis.b16.conv0.affine.weight", "synthesis.generator.tri_plane_synthesis.b16.conv0.affine.bias", "synthesis.generator.tri_plane_synthesis.b16.conv1.weight", "synthesis.generator.tri_plane_synthesis.b16.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b16.conv1.bias", "synthesis.generator.tri_plane_synthesis.b16.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b16.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b16.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b16.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b16.totex.weight", "synthesis.generator.tri_plane_synthesis.b16.totex.bias", "synthesis.generator.tri_plane_synthesis.b16.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b16.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b16.togeo.weight", "synthesis.generator.tri_plane_synthesis.b16.togeo.bias", "synthesis.generator.tri_plane_synthesis.b16.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b16.togeo.affine.bias", "synthesis.generator.tri_plane_synthesis.b32.resample_filter", "synthesis.generator.tri_plane_synthesis.b32.conv0.weight", "synthesis.generator.tri_plane_synthesis.b32.conv0.noise_strength", "synthesis.generator.tri_plane_synthesis.b32.conv0.bias", "synthesis.generator.tri_plane_synthesis.b32.conv0.resample_filter", "synthesis.generator.tri_plane_synthesis.b32.conv0.noise_const", "synthesis.generator.tri_plane_synthesis.b32.conv0.affine.weight", "synthesis.generator.tri_plane_synthesis.b32.conv0.affine.bias", "synthesis.generator.tri_plane_synthesis.b32.conv1.weight", "synthesis.generator.tri_plane_synthesis.b32.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b32.conv1.bias", "synthesis.generator.tri_plane_synthesis.b32.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b32.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b32.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b32.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b32.totex.weight", "synthesis.generator.tri_plane_synthesis.b32.totex.bias", "synthesis.generator.tri_plane_synthesis.b32.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b32.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b32.togeo.weight", "synthesis.generator.tri_plane_synthesis.b32.togeo.bias", "synthesis.generator.tri_plane_synthesis.b32.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b32.togeo.affine.bias", "synthesis.generator.tri_plane_synthesis.b64.resample_filter", "synthesis.generator.tri_plane_synthesis.b64.conv0.weight", "synthesis.generator.tri_plane_synthesis.b64.conv0.noise_strength", "synthesis.generator.tri_plane_synthesis.b64.conv0.bias", "synthesis.generator.tri_plane_synthesis.b64.conv0.resample_filter", "synthesis.generator.tri_plane_synthesis.b64.conv0.noise_const", "synthesis.generator.tri_plane_synthesis.b64.conv0.affine.weight", "synthesis.generator.tri_plane_synthesis.b64.conv0.affine.bias", "synthesis.generator.tri_plane_synthesis.b64.conv1.weight", "synthesis.generator.tri_plane_synthesis.b64.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b64.conv1.bias", "synthesis.generator.tri_plane_synthesis.b64.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b64.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b64.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b64.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b64.totex.weight", "synthesis.generator.tri_plane_synthesis.b64.totex.bias", "synthesis.generator.tri_plane_synthesis.b64.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b64.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b64.togeo.weight", "synthesis.generator.tri_plane_synthesis.b64.togeo.bias", "synthesis.generator.tri_plane_synthesis.b64.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b64.togeo.affine.bias", "synthesis.generator.tri_plane_synthesis.b128.resample_filter", "synthesis.generator.tri_plane_synthesis.b128.conv0.weight", "synthesis.generator.tri_plane_synthesis.b128.conv0.noise_strength", "synthesis.generator.tri_plane_synthesis.b128.conv0.bias", "synthesis.generator.tri_plane_synthesis.b128.conv0.resample_filter", "synthesis.generator.tri_plane_synthesis.b128.conv0.noise_const", "synthesis.generator.tri_plane_synthesis.b128.conv0.affine.weight", "synthesis.generator.tri_plane_synthesis.b128.conv0.affine.bias", "synthesis.generator.tri_plane_synthesis.b128.conv1.weight", "synthesis.generator.tri_plane_synthesis.b128.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b128.conv1.bias", "synthesis.generator.tri_plane_synthesis.b128.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b128.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b128.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b128.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b128.totex.weight", "synthesis.generator.tri_plane_synthesis.b128.totex.bias", "synthesis.generator.tri_plane_synthesis.b128.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b128.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b128.togeo.weight", "synthesis.generator.tri_plane_synthesis.b128.togeo.bias", "synthesis.generator.tri_plane_synthesis.b128.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b128.togeo.affine.bias", "synthesis.generator.tri_plane_synthesis.b256.resample_filter", "synthesis.generator.tri_plane_synthesis.b256.conv0.weight", "synthesis.generator.tri_plane_synthesis.b256.conv0.noise_strength", "synthesis.generator.tri_plane_synthesis.b256.conv0.bias", "synthesis.generator.tri_plane_synthesis.b256.conv0.resample_filter", "synthesis.generator.tri_plane_synthesis.b256.conv0.noise_const", "synthesis.generator.tri_plane_synthesis.b256.conv0.affine.weight", "synthesis.generator.tri_plane_synthesis.b256.conv0.affine.bias", "synthesis.generator.tri_plane_synthesis.b256.conv1.weight", "synthesis.generator.tri_plane_synthesis.b256.conv1.noise_strength", "synthesis.generator.tri_plane_synthesis.b256.conv1.bias", "synthesis.generator.tri_plane_synthesis.b256.conv1.resample_filter", "synthesis.generator.tri_plane_synthesis.b256.conv1.noise_const", "synthesis.generator.tri_plane_synthesis.b256.conv1.affine.weight", "synthesis.generator.tri_plane_synthesis.b256.conv1.affine.bias", "synthesis.generator.tri_plane_synthesis.b256.totex.weight", "synthesis.generator.tri_plane_synthesis.b256.totex.bias", "synthesis.generator.tri_plane_synthesis.b256.totex.affine.weight", "synthesis.generator.tri_plane_synthesis.b256.totex.affine.bias", "synthesis.generator.tri_plane_synthesis.b256.togeo.weight", "synthesis.generator.tri_plane_synthesis.b256.togeo.bias", "synthesis.generator.tri_plane_synthesis.b256.togeo.affine.weight", "synthesis.generator.tri_plane_synthesis.b256.togeo.affine.bias", "synthesis.generator.mlp_synthesis_tex.layers.0.weight", "synthesis.generator.mlp_synthesis_tex.layers.0.bias", "synthesis.generator.mlp_synthesis_tex.layers.0.affine.weight", "synthesis.generator.mlp_synthesis_tex.layers.0.affine.bias", "synthesis.generator.mlp_synthesis_tex.layers.1.weight", "synthesis.generator.mlp_synthesis_tex.layers.1.bias", "synthesis.generator.mlp_synthesis_tex.layers.1.affine.weight", "synthesis.generator.mlp_synthesis_tex.layers.1.affine.bias", "synthesis.generator.mlp_synthesis_sdf.layers.0.weight", "synthesis.generator.mlp_synthesis_sdf.layers.0.bias", "synthesis.generator.mlp_synthesis_sdf.layers.0.affine.weight", "synthesis.generator.mlp_synthesis_sdf.layers.0.affine.bias", "synthesis.generator.mlp_synthesis_sdf.layers.1.weight", "synthesis.generator.mlp_synthesis_sdf.layers.1.bias", "synthesis.generator.mlp_synthesis_sdf.layers.1.affine.weight", "synthesis.generator.mlp_synthesis_sdf.layers.1.affine.bias", "synthesis.generator.mlp_synthesis_def.layers.0.weight", "synthesis.generator.mlp_synthesis_def.layers.0.bias", "synthesis.generator.mlp_synthesis_def.layers.0.affine.weight", "synthesis.generator.mlp_synthesis_def.layers.0.affine.bias", "synthesis.generator.mlp_synthesis_def.layers.1.weight", "synthesis.generator.mlp_synthesis_def.layers.1.bias", "synthesis.generator.mlp_synthesis_def.layers.1.affine.weight", "synthesis.generator.mlp_synthesis_def.layers.1.affine.bias". 
        Unexpected key(s) in state_dict: "synthesis.geometry_synthesis_sdf.b4.const", "synthesis.geometry_synthesis_sdf.b4.conv1.weight", "synthesis.geometry_synthesis_sdf.b4.conv1.noise_strength", "synthesis.geometry_synthesis_sdf.b4.conv1.bias", "synthesis.geometry_synthesis_sdf.b4.conv1.noise_const", "synthesis.geometry_synthesis_sdf.b4.conv1.affine.weight", "synthesis.geometry_synthesis_sdf.b4.conv1.affine.bias", "synthesis.geometry_synthesis_sdf.b8.conv0.weight", "synthesis.geometry_synthesis_sdf.b8.conv0.noise_strength", "synthesis.geometry_synthesis_sdf.b8.conv0.bias", "synthesis.geometry_synthesis_sdf.b8.conv0.noise_const", "synthesis.geometry_synthesis_sdf.b8.conv0.affine.weight", "synthesis.geometry_synthesis_sdf.b8.conv0.affine.bias", "synthesis.geometry_synthesis_sdf.b8.conv1.weight", "synthesis.geometry_synthesis_sdf.b8.conv1.noise_strength", "synthesis.geometry_synthesis_sdf.b8.conv1.bias", "synthesis.geometry_synthesis_sdf.b8.conv1.noise_const", "synthesis.geometry_synthesis_sdf.b8.conv1.affine.weight", "synthesis.geometry_synthesis_sdf.b8.conv1.affine.bias", "synthesis.geometry_synthesis_sdf.b8.skip.weight", "synthesis.geometry_synthesis_sdf.b8.skip.resample_filter", "synthesis.geometry_synthesis_sdf.b16.conv0.weight", "synthesis.geometry_synthesis_sdf.b16.conv0.noise_strength", "synthesis.geometry_synthesis_sdf.b16.conv0.bias", "synthesis.geometry_synthesis_sdf.b16.conv0.noise_const", "synthesis.geometry_synthesis_sdf.b16.conv0.affine.weight", "synthesis.geometry_synthesis_sdf.b16.conv0.affine.bias", "synthesis.geometry_synthesis_sdf.b16.conv1.weight", "synthesis.geometry_synthesis_sdf.b16.conv1.noise_strength", "synthesis.geometry_synthesis_sdf.b16.conv1.bias", "synthesis.geometry_synthesis_sdf.b16.conv1.noise_const", "synthesis.geometry_synthesis_sdf.b16.conv1.affine.weight", "synthesis.geometry_synthesis_sdf.b16.conv1.affine.bias", "synthesis.geometry_synthesis_sdf.b16.skip.weight", "synthesis.geometry_synthesis_sdf.b16.skip.resample_filter", "synthesis.geometry_synthesis_sdf.b32.conv0.weight", "synthesis.geometry_synthesis_sdf.b32.conv0.noise_strength", "synthesis.geometry_synthesis_sdf.b32.conv0.bias", "synthesis.geometry_synthesis_sdf.b32.conv0.noise_const", "synthesis.geometry_synthesis_sdf.b32.conv0.affine.weight", "synthesis.geometry_synthesis_sdf.b32.conv0.affine.bias", "synthesis.geometry_synthesis_sdf.b32.conv1.weight", "synthesis.geometry_synthesis_sdf.b32.conv1.noise_strength", "synthesis.geometry_synthesis_sdf.b32.conv1.bias", "synthesis.geometry_synthesis_sdf.b32.conv1.noise_const", "synthesis.geometry_synthesis_sdf.b32.conv1.affine.weight", "synthesis.geometry_synthesis_sdf.b32.conv1.affine.bias", "synthesis.geometry_synthesis_sdf.b32.skip.weight", "synthesis.geometry_synthesis_sdf.b32.skip.resample_filter", "synthesis.geometry_synthesis_sdf.layers.0.weight", "synthesis.geometry_synthesis_sdf.layers.0.bias", "synthesis.geometry_synthesis_sdf.layers.0.affine.weight", "synthesis.geometry_synthesis_sdf.layers.0.affine.bias", "synthesis.geometry_synthesis_sdf.layers.1.weight", "synthesis.geometry_synthesis_sdf.layers.1.bias", "synthesis.geometry_synthesis_sdf.layers.1.affine.weight", "synthesis.geometry_synthesis_sdf.layers.1.affine.bias", "synthesis.geometry_synthesis_sdf.layers.2.weight", "synthesis.geometry_synthesis_sdf.layers.2.bias", "synthesis.geometry_synthesis_sdf.layers.2.affine.weight", "synthesis.geometry_synthesis_sdf.layers.2.affine.bias", "synthesis.geometry_synthesis_sdf.layers.3.weight", "synthesis.geometry_synthesis_sdf.layers.3.bias", "synthesis.geometry_synthesis_sdf.layers.3.affine.weight", "synthesis.geometry_synthesis_sdf.layers.3.affine.bias", "synthesis.geometry_synthesis_def.b4.const", "synthesis.geometry_synthesis_def.b4.conv1.weight", "synthesis.geometry_synthesis_def.b4.conv1.noise_strength", "synthesis.geometry_synthesis_def.b4.conv1.bias", "synthesis.geometry_synthesis_def.b4.conv1.noise_const", "synthesis.geometry_synthesis_def.b4.conv1.affine.weight", "synthesis.geometry_synthesis_def.b4.conv1.affine.bias", "synthesis.geometry_synthesis_def.b8.conv0.weight", "synthesis.geometry_synthesis_def.b8.conv0.noise_strength", "synthesis.geometry_synthesis_def.b8.conv0.bias", "synthesis.geometry_synthesis_def.b8.conv0.noise_const", "synthesis.geometry_synthesis_def.b8.conv0.affine.weight", "synthesis.geometry_synthesis_def.b8.conv0.affine.bias", "synthesis.geometry_synthesis_def.b8.conv1.weight", "synthesis.geometry_synthesis_def.b8.conv1.noise_strength", "synthesis.geometry_synthesis_def.b8.conv1.bias", "synthesis.geometry_synthesis_def.b8.conv1.noise_const", "synthesis.geometry_synthesis_def.b8.conv1.affine.weight", "synthesis.geometry_synthesis_def.b8.conv1.affine.bias", "synthesis.geometry_synthesis_def.b8.skip.weight", "synthesis.geometry_synthesis_def.b8.skip.resample_filter", "synthesis.geometry_synthesis_def.b16.conv0.weight", "synthesis.geometry_synthesis_def.b16.conv0.noise_strength", "synthesis.geometry_synthesis_def.b16.conv0.bias", "synthesis.geometry_synthesis_def.b16.conv0.noise_const", "synthesis.geometry_synthesis_def.b16.conv0.affine.weight", "synthesis.geometry_synthesis_def.b16.conv0.affine.bias", "synthesis.geometry_synthesis_def.b16.conv1.weight", "synthesis.geometry_synthesis_def.b16.conv1.noise_strength", "synthesis.geometry_synthesis_def.b16.conv1.bias", "synthesis.geometry_synthesis_def.b16.conv1.noise_const", "synthesis.geometry_synthesis_def.b16.conv1.affine.weight", "synthesis.geometry_synthesis_def.b16.conv1.affine.bias", "synthesis.geometry_synthesis_def.b16.skip.weight", "synthesis.geometry_synthesis_def.b16.skip.resample_filter", "synthesis.geometry_synthesis_def.b32.conv0.weight", "synthesis.geometry_synthesis_def.b32.conv0.noise_strength", "synthesis.geometry_synthesis_def.b32.conv0.bias", "synthesis.geometry_synthesis_def.b32.conv0.noise_const", "synthesis.geometry_synthesis_def.b32.conv0.affine.weight", "synthesis.geometry_synthesis_def.b32.conv0.affine.bias", "synthesis.geometry_synthesis_def.b32.conv1.weight", "synthesis.geometry_synthesis_def.b32.conv1.noise_strength", "synthesis.geometry_synthesis_def.b32.conv1.bias", "synthesis.geometry_synthesis_def.b32.conv1.noise_const", "synthesis.geometry_synthesis_def.b32.conv1.affine.weight", "synthesis.geometry_synthesis_def.b32.conv1.affine.bias", "synthesis.geometry_synthesis_def.b32.skip.weight", "synthesis.geometry_synthesis_def.b32.skip.resample_filter", "synthesis.geometry_synthesis_def.layers.0.weight", "synthesis.geometry_synthesis_def.layers.0.bias", "synthesis.geometry_synthesis_def.layers.0.affine.weight", "synthesis.geometry_synthesis_def.layers.0.affine.bias", "synthesis.geometry_synthesis_def.layers.1.weight", "synthesis.geometry_synthesis_def.layers.1.bias", "synthesis.geometry_synthesis_def.layers.1.affine.weight", "synthesis.geometry_synthesis_def.layers.1.affine.bias", "synthesis.geometry_synthesis_def.layers.2.weight", "synthesis.geometry_synthesis_def.layers.2.bias", "synthesis.geometry_synthesis_def.layers.2.affine.weight", "synthesis.geometry_synthesis_def.layers.2.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b4.torgb.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv0.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b8.torgb.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv0.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b16.torgb.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv0.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b32.torgb.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv0.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b64.torgb.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv0.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b128.torgb.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv0.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.noise_strength", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.resample_filter", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.noise_const", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.conv1.affine.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.torgb.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.torgb.bias", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.torgb.affine.weight", "synthesis.geometry_synthesis_tex.tri_plane_synthesis.b256.torgb.affine.bias", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.0.weight", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.0.bias", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.0.affine.weight", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.0.affine.bias", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.1.weight", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.1.bias", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.1.affine.weight", "synthesis.geometry_synthesis_tex.mlp_synthesis.layers.1.affine.bias".

training on FFHQ?

Thanks for your great work. I find that you compare your result with eg3d in paper and claim that your result is better than it in 3 dataset. I have a question that GET3D is better than EG3D in human face dataset like FFHQ or CelebA.

Ability to reduce the video memory requirements for the result output

I have ~~a dream...~~ my good old 3060 with 12 GB of memory, and this project needs all 16 to get results. Is it possible to reduce its appetite by a measly 4 gigabytes?

The camera intrinsic of the rendered data

Hi, thanks for your great work.
I am trying to find the camera intrinsic from the rendering scripts but I did not find anything related to it.
Could you please give me some hint about how to determine the intrinsic of the camera in your rendering scripts?

question about setting blender

Hi, @SteveJunGao ,

According to the steps, the following two lines are just used for ensuring the existence of pip and numpy:

./python3.7m -m ensurepip
./python3.7m -m pip install numpy

Do we need to specify the location of bpy module, as it is needed here, when executing render_all.py ?

When I execute

python render_all.py --save_folder /media/root/data2/ShapeNet/render_view --dataset_folder /media/root/data2/ShapeNet/ShapeNetCore.v1 --blender_root /root/Downloads/blender-2.90.0-linux64/2.90/python/bin

I got the following error:

sh: 1: /root/Downloads/blender-2.90.0-linux64/2.90/python/bin: Permission denied

Any hints to solve this issue?

Thanks~

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.