princeton-vl / cornernet-lite Goto Github PK

Excuse me?
Where is the GPU setup?

Running shows that GPU = None, in fact, I have a gpu, cuda10 installation training is no problem. thank you.

请问一下，gpu在哪里设置？运行显示 gpu=None，实际我gpu是有的，cuda10都安装训练没问题。方便加QQ沟通一下吗？ QQ2737499951,谢谢

================
args= Namespace(cfg_file='CornerNet_Saccade', dist_backend='nccl', dist_url=None, distributed=False, gpu=None, initialize=False, rank=0, start_iter=0, workers=2, world_size=-1)
train tart_iter 0 distributed False world_size -1 initialize False gpu None
Process 0: 创建模型 building model...
total parameters: 116969339
启动预取数据 start prefetching data...

An error occurred during training Can you show your dataset tree directory? Thank you (the following is my structure according to the document)
---CornerNet-Lite
---data
---coco
--- images
---trainval2014
---train2014.zip
---minival2014
---val2014.zip
---testdev2017
---test2017.zip

run python train.py CornerNet_Saccade --workers=1

File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'
Process Process-1:
Traceback (most recent call last):
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
self.run()
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 99, in run
self._target(*self._args, **self._kwargs)
File "train.py", line 59, in prefetch_data
raise e
File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'
setting learning rate to: 0.00025
training start...
start prefetching data...
shuffling indices...
Traceback (most recent call last):
File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'
Process Process-2:
Traceback (most recent call last):
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
self.run()
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 99, in run
self._target(*self._args, **self._kwargs)
File "train.py", line 59, in prefetch_data
raise e
File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'

This error occurs when I run demo.py. Is this a problem with pytorch? However, the environment is executed according to the configuration file. thank you

loading from /test/CornerNet-Lite/core/../cache/nnet/CornerNet_Saccade/CornerNet_Saccade_500000.pkl
/opt/conda/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/nn/functional.py:2423: UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details.
"See the documentation of nn.Upsample for details.".format(mode))

Hi @heilaw
I use default input size 511 to train and test, test_scales is 1 and test_flipped is false.
When I evaluate, I find input image size is the original image size not the input size, is that correct? I think input image should be prepocessed to 512 before input into network to do inference.
resize

./data/ch/val/JPEGImages/10016d000d0d6e0e8.jpg
torch.Size([3, 511, 767])
./data/ch/val/JPEGImages/1028c8000e8c388b9.jpg
torch.Size([3, 1023, 1407])
./data/ch/val/JPEGImages/1010cc000f912334d.jpg
torch.Size([3, 767, 1151])
./data/ch/val/JPEGImages/10034e000b214be47.jpg
torch.Size([3, 1023, 1407])
./data/ch/val/JPEGImages/101ae50008d3ac841.jpg
torch.Size([3, 4351, 3327])

it is clear that the bound boxes' size are wrongly identified for that the cornerpoints are wrongly grouped.
is there any ways such as nms configurations rather than modifying networks to avoid such kind of mistakenly predicted boxes?

Because of the lack of compution resource, so I have to train it on cpu,is it feasible?What part of code should I change?

Thank you for your open source, but I can't use the provided trained model to test, and the following error occurred:RuntimeError: Error(s) in loading state_dict for DummyModule.

First Image is the CornerNet-Squeeze result , second Image is yolo v3 result

I get more than 200.0s/it.
It`s too slow for training a detection model.

run demo.py
after "loading from /content/ai/hdy_conornet/CornerNet-Lite/core/../cache/nnet/CornerNet_Squeeze/CornerNet_Squeeze_500000.pkl"

用来训练自己的数据，启动后发现CPU占用很高，GPU占用仅仅50%左右

I want to train my own dataset.
I see you use the eigenvalue and eigenvector during data sampling.
Could you tell me the method you use in details ?

Right now I only want to try to infer to compare with other models but I don't have a gpu in my computer, how I should change the code to make it?

I couldn't find where the ground truth code input into the model. And ground truth shape is?

File "./code/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "./code/CornerNet-Lite/core/sample/cornernet.py", line 139, in cornernet
tl_regrs[b_ind, tag_ind, :] = [fxtl - xtl, fytl - ytl]
IndexError: index 128 is out of bounds for axis 1 with size 128

Solution:
bounding boxes number of current image exceed 128
set max_tag_len in cornetnet.py = 500 or max number of boxes in images

Hey, thanks for your work.
I want to try your net, I did create the conda environment. But when I execute python demo.py:

I only have one GPU, RTX2080.
Waiting for your suggestions.

I want to tran the model for my dataset which has 9 categories.
I just modify the coco.py for the data. Beside I set the CornerNet_Saccade.py

	#TODO 80_>9
        tl_heats = nn.ModuleList([self._pred_mod(9) for _ in range(stacks)])
        br_heats = nn.ModuleList([self._pred_mod(9) for _ in range(stacks)])

So how to use your pretrained model for the train process ?

I change input image size to 800,800 in configs/CornerNet_Saccade.json,but it is wrong.

Hi,
since it took so long to train (around two and a half week to train with 4 1080Ti GPUs.). is there any tool to monitor the training process?

Traceback (most recent call last):
File "train.py", line 253, in
main(1, ngpus_per_node, args)
File "train.py", line 236, in main
train(training_dbs, validation_db, system_config, model, args)
File "train.py", line 186, in train
nnet.set_lr(learning_rate)
File "/usr/lib/python3.5/contextlib.py", line 77, in exit
self.gen.throw(type, value, traceback)
File "/data/mj/CornerNet-Lite/core/utils/tqdm.py", line 23, in stdout_to_tqdm
raise exc
File "/data/mj/CornerNet-Lite/core/utils/tqdm.py", line 21, in stdout_to_tqdm
yield save_stdout
File "train.py", line 168, in train
training_loss = nnet.train(**training)
File "/data/mj/CornerNet-Lite/core/nnet/py_factory.py", line 93, in train
loss = self.network(xs, ys)
File "/home/mj/.local/lib/python3.5/site-packages/torch/nn/modules/module.py", line 489, in call
result = self.forward(*input, **kwargs)
File "/data/mj/CornerNet-Lite/core/models/py_utils/data_parallel.py", line 66, in forward
inputs, kwargs = self.scatter(inputs, kwargs, self.device_ids, self.chunk_sizes)
File "/data/mj/CornerNet-Lite/core/models/py_utils/data_parallel.py", line 77, in scatter
return scatter_kwargs(inputs, kwargs, device_ids, dim=self.dim, chunk_sizes=self.chunk_sizes)
File "/data/mj/CornerNet-Lite/core/models/py_utils/scatter_gather.py", line 30, in scatter_kwargs
inputs = scatter(inputs, target_gpus, dim, chunk_sizes) if inputs else []
File "/data/mj/CornerNet-Lite/core/models/py_utils/scatter_gather.py", line 25, in scatter
return scatter_map(inputs)
File "/data/mj/CornerNet-Lite/core/models/py_utils/scatter_gather.py", line 18, in scatter_map
return list(zip(map(scatter_map, obj)))
File "/data/mj/CornerNet-Lite/core/models/py_utils/scatter_gather.py", line 20, in scatter_map
return list(map(list, zip(map(scatter_map, obj))))
File "/data/mj/CornerNet-Lite/core/models/py_utils/scatter_gather.py", line 15, in scatter_map
return Scatter.apply(target_gpus, chunk_sizes, dim, obj)
File "/home/mj/.local/lib/python3.5/site-packages/torch/nn/parallel/_functions.py", line 89, in forward
outputs = comm.scatter(input, target_gpus, chunk_sizes, ctx.dim, streams)
File "/home/mj/.local/lib/python3.5/site-packages/torch/cuda/comm.py", line 148, in scatter
return tuple(torch._C._scatter(tensor, devices, chunk_sizes, dim, streams))
RuntimeError: CUDA error: invalid device ordinal (exchangeDevice at /pytorch/aten/src/ATen/cuda/detail/CUDAGuardImpl.h:28)
frame #0: std::function<std::string ()>::operator()() const + 0x11 (0x7f08e8ee5021 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libc10.so)
frame #1: c10::Error::Error(c10::SourceLocation, std::string const&) + 0x2a (0x7f08e8ee48ea in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libc10.so)
frame #2: + 0x4e426f (0x7f09235f526f in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libtorch_python.so)
frame #3: + 0x8cdfa2 (0x7f08e99c3fa2 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libcaffe2.so)
frame #4: + 0xa14ae5 (0x7f08e9b0aae5 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libcaffe2.so)
frame #5: at::TypeDefault::copy(at::Tensor const&, bool, c10::optionalc10::Device) const + 0x56 (0x7f08e9c47c76 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libcaffe2.so)
frame #6: + 0x977f47 (0x7f08e9a6df47 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libcaffe2.so)
frame #7: at::native::to(at::Tensor const&, at::TensorOptions const&, bool, bool) + 0x295 (0x7f08e9a6faf5 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libcaffe2.so)
frame #8: at::TypeDefault::to(at::Tensor const&, at::TensorOptions const&, bool, bool) const + 0x17 (0x7f08e9c0e4f7 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libcaffe2.so)
frame #9: torch::autograd::VariableType::to(at::Tensor const&, at::TensorOptions const&, bool, bool) const + 0x17a (0x7f08e814ebaa in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libtorch.so.1)
frame #10: torch::cuda::scatter(at::Tensor const&, c10::ArrayRef, c10::optional<std::vector<long, std::allocator > > const&, long, c10::optional<std::vector<c10::optionalat::cuda::CUDAStream, std::allocator<c10::optionalat::cuda::CUDAStream > > > const&) + 0x391 (0x7f09235f75f1 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libtorch_python.so)
frame #11: + 0x4ebd4f (0x7f09235fcd4f in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libtorch_python.so)
frame #12: + 0x11642e (0x7f092322742e in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libtorch_python.so)

frame #15: python() [0x53fc97]
frame #18: python() [0x4ec2e3]
frame #21: THPFunction_apply(_object, _object) + 0x581 (0x7f0923423ab1 in /home/mj/.local/lib/python3.5/site-packages/torch/lib/libtorch_python.so)
frame #25: python() [0x4ec2e3]
frame #27: python() [0x535f0e]
frame #32: python() [0x4ec2e3]
frame #34: python() [0x535f0e]
frame #38: python() [0x5401ef]
frame #40: python() [0x5401ef]
frame #42: python() [0x53fc97]
frame #46: python() [0x4ec3f7]
frame #50: python() [0x4ec2e3]
frame #52: python() [0x4fbfce]
frame #54: python() [0x574db6]
frame #58: python() [0x4ec3f7]
frame #62: python() [0x5401ef]

what is wrong?

Hi,
I noticed that in the paper you mentioned the model can be trained with 4 GTX 1080-Ti.
I only have 2 GTX-1080-Ti, can I train it?
Thanks and best regards,

Liming

Based on CornerNet, detects each object as a triplet, rather than a pair, of keypoints, CenterNet has made great progress. Do you have a test comparison between CenterNet and CornerNet-Lite? Thank you！

https://github.com/Duankaiwen/CenterNet

CenterNet: Keypoint Triplets for Object Detection CenterNet：用于对象检测的关键点三元组

CenterNet is an one-stage detector which gets trained from scratch. On the MS-COCO dataset, CenterNet achieves an AP of 47.0%, which surpasses all known one-stage detectors, and even gets very close to the top-performance two-stage detectors.

Description of their paper: “We build our framework upon a representative one-stage keypoint-based detector named CornerNet. Our approach, named CenterNet, detects each object as a triplet, rather than a pair, of keypoints, which improves both precision and recall. Accordingly, we design two customized modules named cascade corner pooling and center pooling, which play the roles of enriching information collected by both top-left and bottom-right corners and providing more recognizable information at the central regions, respectively.”

Hi, author!
I wanna ask some questions about the training speed of your CornerNet_Squeeze with the default hyper-parameters in your configuration.
How long did you take to get the results (on the coco dataset) in the paper compared with yolo_v3?

In losses.py, when calcalating ae_loss, the push loss mentioned as:

dist = dist - 1 / (num + 1e-4)

I couldn't understand it? If you want to subtract the j=k in the formula, may be not use this method?
Looking forward your reply. thks

when run python demo.py, I met this question:
`python demo.py
total parameters: 116969339
loading from /home/qing/ngs/CornerNet-Lite/core/../cache/nnet/CornerNet_Saccade/CornerNet_Saccade_500000.pkl

Traceback (most recent call last):
File "demo.py", line 7, in
detector = CornerNet_Saccade()
File "/home/qing/ngs/CornerNet-Lite/core/detectors.py", line 49, in init
super(CornerNet_Saccade, self).init(coco, cornernet, cornernet_saccade_inference, model=model_path)
File "/home/qing/ngs/CornerNet-Lite/core/base.py", line 14, in init
self._nnet.load_pretrained_params(model)
File "/home/qing/ngs/CornerNet-Lite/core/nnet/py_factory.py", line 123, in load_pretrained_params
params = torch.load(f)
File "/home/qing/anaconda3/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/serialization.py", line 367, in load
return _load(f, map_location, pickle_module)
File "/home/qing/anaconda3/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/serialization.py", line 538, in _load
result = unpickler.load()
File "/home/qing/anaconda3/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/serialization.py", line 504, in persistent_load
data_type(size), location)
File "/home/qing/anaconda3/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/serialization.py", line 113, in default_restore_location
result = fn(storage, location)
File "/home/qing/anaconda3/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/serialization.py", line 94, in _cuda_deserialize
device = validate_cuda_device(location)
File "/home/qing/anaconda3/envs/CornerNet_Lite/lib/python3.7/site-packages/torch/serialization.py", line 78, in validate_cuda_device
raise RuntimeError('Attempting to deserialize object on a CUDA '
RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location='cpu' to map your storages to the CPU.
`

CUDA Version 8.0.44

I have 4 Nvidia P40 GPUs

An error occurred during training Can you show your dataset tree directory? Thank you (the following is my structure according to the document)
---CornerNet-Lite
------data
----------coco
---------------images
---------------------trainval2014
---------------------------------train2014.zip
---------------------minival2014
---------------------------------val2014.zip
---------------------testdev2017
---------------------------------test2017.zip

run python train.py CornerNet_Saccade --workers=1

File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'
Process Process-1:
Traceback (most recent call last):
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
self.run()
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 99, in run
self._target(*self._args, **self._kwargs)
File "train.py", line 59, in prefetch_data
raise e
File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'
setting learning rate to: 0.00025
training start...
start prefetching data...
shuffling indices...
Traceback (most recent call last):
File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'
Process Process-2:
Traceback (most recent call last):
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
self.run()
File "/opt/conda/envs/CornerNet_Lite/lib/python3.7/multiprocessing/process.py", line 99, in run
self._target(*self._args, **self._kwargs)
File "train.py", line 59, in prefetch_data
raise e
File "train.py", line 55, in prefetch_data
data, ind = sample_data(system_config, db, ind, data_aug=data_aug)
File "/test/CornerNet-Lite/core/sample/init.py", line 5, in data_sampling_func
return globals()[sys_configs.sampling_function](sys_configs, db, k_ind, data_aug, debug)
File "/test/CornerNet-Lite/core/sample/cornernet_saccade.py", line 169, in cornernet_saccade
border = [0, image.shape[0], 0, image.shape[1]]
AttributeError: 'NoneType' object has no attribute 'shape'

What is the attention map? Is it the same as a feature map? What is the gt attention?

CornerNet-Lite数据集
链接：https://pan.baidu.com/s/141G-JZuF2EHypJwIgbicbw
提取码：7z3k

**CornerNet_Squeeze_500000.pkl (122M)
CornerNet_Saccade_500000.pkl (447M)
CornerNet_500000.pkl (768M)
annotations.zip(157.5M)
**

Put the CornerNet-Saccade model under /cache/nnet/CornerNet_Saccade/, CornerNet-Squeeze model under /cache/nnet/CornerNet_Squeeze/ and CornerNet model under /cache/nnet/CornerNet/. (* Note we use underscore instead of dash in both the directory names for CornerNet-Saccade and CornerNet-Squeeze.)

==========
将CornerNet-Saccade模型放在下面/cache/nnet/CornerNet_Saccade/，将CornerNet-Squeeze模型放在下面/cache/nnet/CornerNet_Squeeze/，使用CornerNet模型/cache/nnet/CornerNet/。（*注意我们在CornerNet-Saccade和CornerNet-Squeeze的目录名中使用下划线而不是破折号。）

注意：CornerNet模型与原始CornerNet存储库中的模型相同。我们刚将它移植到这个新的仓库。

请问下作者，１．为什末两个阶段要共享一个模型和参数呢？
２．预测时先用255*255的图预测location，再使用location的高清图输入网络，训练时是如何保证这两类图片都有被喂给网络呢？

and i want to visualize the network ,how to do it?? can you help me

First Image is CornerNet-Saccade result and second Image is CornerNet-squeeze result.

/data/coco/images/trainval2014 ===========>改用/data/coco/images/train2017
/data/coco/images/minival2014 ===========>改用/data/coco/images/val2017
/data/coco/images/testdev2017 ===========>改用/data/coco/images/test2017

/data/coco/annotations/instances_trainval2014.json ======>/data/coco/annotations/instances_train2017.json
/data/coco/annotations/instances_minival2014.json ===========>/data/coco/annotations/instances_val2017.json
/data/coco/annotations/instances_testdev2017.json ===========>/data/coco/annotations/instances_testdev2017.json

===========
D:\Tensorflow\pytorch\CornerNet-Lite>python train.py CornerNet_Saccade
Process 0: loading all datasets...
Process 0: using 4 workers
Traceback (most recent call last):
File "train.py", line 249, in
main(None, ngpus_per_node, args)
File "train.py", line 220, in main
training_dbs = [datasets[dataset](config["db"], split=train_split, sys_config=system_config) for _ in range(workers)]
File "train.py", line 220, in
training_dbs = [datasets[dataset](config["db"], split=train_split, sys_config=system_config) for _ in range(workers)]
File "D:\Tensorflow\pytorch\CornerNet-Lite\core\dbs\coco.py", line 67, in init
}[split]
KeyError: 'trainval'

调用这些函数时出现未定义的符号，求解答

I want to use CornerNet_Squeeze.pkl on Mobile Phone of Android or IOS .
Could you tell me the method ?

0.4s on video

when I run the python demo.py , OOM occurred like below:
RuntimeError: CUDA out of memory. Tried to allocate 44.00 MiB (GPU 0; 1.95 GiB total capacity; 1.23 GiB already allocated; 2.62 MiB free; 98.86 MiB cached)

so where is bitch_size?

thx ^_^

Any speed enhancement in lite version?

Hello,
Thanks a lot for the excellent job from Princeton-VL, our system is based on the Nvidia-GPU driver of CUDA-9.0.
Can I just directly try the open source code on my system with the configuration file change.
Or should I upgrade my driver?
Thanks a lot!

Why I cannot download modules？ The link is not valid？

Hi,

Does anybody have such issue when comiling core/models/py_utils/_cpools?
CornerNet_Lite) lhu@LAB00100W:/mnt/hdd1_6tb/experiment/CornerNet_Lite/core/models/py_utils/_cpools$ python setup.py install --user
running install
running bdist_egg
running egg_info
writing cpools.egg-info/PKG-INFO
writing dependency_links to cpools.egg-info/dependency_links.txt
writing top-level names to cpools.egg-info/top_level.txt
reading manifest file 'cpools.egg-info/SOURCES.txt'
writing manifest file 'cpools.egg-info/SOURCES.txt'
installing library code to build/bdist.linux-x86_64/egg
running install_lib
running build_ext
building 'top_pool' extension
gcc -pthread -B /home/lhu/anaconda3/envs/CornerNet_Lite/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include -I/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/TH -I/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/THC -I/home/lhu/anaconda3/envs/CornerNet_Lite/include/python3.6m -c src/top_pool.cpp -o build/temp.linux-x86_64-3.6/src/top_pool.o -DTORCH_EXTENSION_NAME=top_pool -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++11
cc1plus: warning: command line option \u2018-Wstrict-prototypes\u2019 is valid for C/ObjC but not for C++
src/top_pool.cpp: In function \u2018std::vectorat::Tensor top_pool_backward(at::Tensor, at::Tensor)\u2019:
src/top_pool.cpp:39:20: error: \u2018zeros\u2019 is not a member of \u2018torch\u2019
auto max_val = torch::zeros({batch, channel, width}, at::device(at::kCUDA).dtype(at::kFloat));
^~~~~
src/top_pool.cpp:39:20: note: suggested alternatives:
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:4374:22: note: \u2018at::zeros\u2019
static inline Tensor zeros(const Type & dtype, IntList size) {
^~~~~
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:11:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/NativeFunctions.h:548:15: note: \u2018at::native::zeros\u2019
AT_API Tensor zeros(IntList size, const TensorOptions & options={});
^~~~~
src/top_pool.cpp:40:20: error: \u2018zeros\u2019 is not a member of \u2018torch\u2019
auto max_ind = torch::zeros({batch, channel, width}, at::device(at::kCUDA).dtype(at::kLong));
^~~~~
src/top_pool.cpp:40:20: note: suggested alternatives:
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:4374:22: note: \u2018at::zeros\u2019
static inline Tensor zeros(const Type & dtype, IntList size) {
^~~~~
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:11:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/NativeFunctions.h:548:15: note: \u2018at::native::zeros\u2019
AT_API Tensor zeros(IntList size, const TensorOptions & options={});
^~~~~
src/top_pool.cpp:52:23: error: \u2018zeros\u2019 is not a member of \u2018torch\u2019
auto gt_mask = torch::zeros({batch, channel, width}, at::device(at::kCUDA).dtype(at::kByte));
^~~~~
src/top_pool.cpp:52:23: note: suggested alternatives:
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:4374:22: note: \u2018at::zeros\u2019
static inline Tensor zeros(const Type & dtype, IntList size) {
^~~~~
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:11:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/NativeFunctions.h:548:15: note: \u2018at::native::zeros\u2019
AT_API Tensor zeros(IntList size, const TensorOptions & options={});
^~~~~
src/top_pool.cpp:53:23: error: \u2018zeros\u2019 is not a member of \u2018torch\u2019
auto max_temp = torch::zeros({batch, channel, width}, at::device(at::kCUDA).dtype(at::kFloat));
^~~~~
src/top_pool.cpp:53:23: note: suggested alternatives:
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:4374:22: note: \u2018at::zeros\u2019
static inline Tensor zeros(const Type & dtype, IntList size) {
^~~~~
In file included from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/Functions.h:11:0,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/ATen.h:14,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/csrc/utils/pybind.h:5,
from /home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/torch/torch.h:6,
from src/top_pool.cpp:1:
/home/lhu/anaconda3/envs/CornerNet_Lite/lib/python3.6/site-packages/torch/lib/include/ATen/NativeFunctions.h:548:15: note: \u2018at::native::zeros\u2019
AT_API Tensor zeros(IntList size, const TensorOptions & options={});
^~~~~
error: command 'gcc' failed with exit status 1
(CornerNet_Lite) lhu@LAB00100W:/mnt/hdd1_6tb/experiment/CornerNet_Lite/core/models/py_utils/_cpools$

I find the push loss is too large in my datasets, it almost 3.1 and difficultly to optimize . Anyone has the same problem?

In core/sample/util.py, there is a funtion called 'gaussian_radius'.

It can dynamic choose the radius of gaussian heatmap.

But I can't understand what does it mean.

Could you pls give me some advices? Thanks a lot.

hello, any one knows the model graph? Which tool could plot it?

Can it support in Windows10?