lmmmeng / transxnet Goto Github PK

Hello authors
thank you for your work!
In the process of reading your paper, your visualization of the effective receptive field in Figure 1 attracted me alot.
Could you please make the visualization code public?
Thanks in advance!

https://github.com/Jacky-Android/TransXNet_recurrent

I run the training of object_detection on Windows, use "python train.py retinanet_transx_t_fpn_1x_coco.py", then I get the mistake:
Traceback (most recent call last):
File "train.py", line 195, in
main()
File "train.py", line 183, in main
train_detector(
File "D:\TransXNet\TransXNet\object_detection\mmdet_custom\apis\train.py", line 184, in train_detector
runner.run(data_loaders, cfg.workflow)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmcv\runner\epoch_based_runner.py", line 136, in run
epoch_runner(data_loaders[i], **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmcv\runner\epoch_based_runner.py", line 53, in train
self.run_iter(data_batch, train_mode=True, **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmcv\runner\epoch_based_runner.py", line 31, in run_iter
outputs = self.model.train_step(data_batch, self.optimizer,
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmcv\parallel\data_parallel.py", line 77, in train_step
return self.module.train_step(*inputs[0], **kwargs[0])
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmdet\models\detectors\base.py", line 248, in train_step
losses = self(**data)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmcv\runner\fp16_utils.py", line 119, in new_func
return old_func(*args, **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmdet\models\detectors\base.py", line 172, in forward
return self.forward_train(img, img_metas, **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmdet\models\detectors\single_stage.py", line 82, in forward_train
x = self.extract_feat(img)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmdet\models\detectors\single_stage.py", line 43, in extract_feat
x = self.backbone(img)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "D:\TransXNet\TransXNet\object_detection\transxnet.py", line 717, in forward
x = self.forward_embeddings(x)
File "D:\TransXNet\TransXNet\object_detection\transxnet.py", line 693, in forward_embeddings
x = self.patch_embed(x)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "D:\TransXNet\TransXNet\object_detection\transxnet.py", line 66, in forward
return self.proj(x)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\mmcv\cnn\bricks\conv_module.py", line 209, in forward
x = self.norm(x)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\nn\modules\module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\nn\modules\batchnorm.py", line 735, in forward
world_size = torch.distributed.get_world_size(process_group)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\distributed\distributed_c10d.py", line 1067, in get_world_size
return _get_group_size(group)
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\distributed\distributed_c10d.py", line 453, in _get_group_size
default_pg = _get_default_group()
File "C:\ProgramData\miniconda3\envs\transxnet\lib\site-packages\torch\distributed\distributed_c10d.py", line 584, in _get_default_group
raise RuntimeError(
RuntimeError: Default process group has not been initialized, please make sure to call init_process_group.

I want to run non-distributed training, please help me.

Hello, can this project only be run or debugged on GPU?

Hi Lou,

   Thanks for your marvelous work TransXNet!
   Recently I am trying to use TransXNet as the encoder to do some dense prediction tasks, I simply do

   **model = TransXNet() or model = transxnet_t()** 

   Then it goes like this

   **TypeError: GELU: __init__() got an unexpected keyword argument 'inplace'**
   
   I am wondering whether this error is related to the version of mmcv.
   Could you tell me how to correctly use TransXNet or kindly provide the requirements.txt ?
   Thanks for your generous reply. :)

May I ask when the code can be open source? Although I am not in this direction, I came across this paper by chance and felt that it was very useful for my direction. I look forward to the author's open source work.

需要弄成coco2017的形式吗？
VOC形式的数据集可以嘛（目标检测）

When I use the Train.py file in my project to train my dataset, I find that the training speed of TransXNet is much slower than models with similar parameter counts, such as PVT_V2, to the point where I believe my server is stuck. I encounter the same issue when using my own training script. Why is this happening? The data is fine because the timm library can train correctly.

您知道这是怎么回事吗 是因为数据集的体量太小了吗