thinksky5124 / svtas Goto Github PK

我建议应该给出一个demo的案例，可以直观的感受可视化

I tried to process the gtea data following the instruction using this script:
python tools/transform_segmentation_label.py data/gtea data/gtea/groundTruth data/gtea --mode localization --fps 15

where can i find the mapping.txt file that should be in the path of gtea?

Hi, thank you for your work.

I am trying to reproduce the paper's results. May i know which is the Transeger configs you've used?

Hi, thank you so much for the excellent contribution and code sharing.

May I inquire whether the extracted I3D features for the EGTEA is available to access?

Thanks!

请问你们asformer在k=32的片段长度下的性能，是使用全视频训练32帧片段测试，还是32帧片段训练32帧片段测试，或是其他的什么训练策略呢？以及是将asformer的block num降低到了多少呢？感谢您的回答

Traceback (most recent call last):
File "main.py", line 121, in
main()
File "main.py", line 111, in main
train(cfg,
File "/home/pengfang/SVTAS-svtas-paper/tasks/train.py", line 202, in train
runner.run_one_iter(data=data, r_tic=r_tic, epoch=epoch)
File "/home/pengfang/SVTAS-svtas-paper/tasks/runner.py", line 259, in run_one_iter
self.run_one_clip(sliding_seg)
File "/home/pengfang/SVTAS-svtas-paper/tasks/runner.py", line 228, in run_one_clip
score, loss_dict = self._model_forward(data_dict)
File "/home/pengfang/SVTAS-svtas-paper/tasks/runner.py", line 200, in _model_forward
outputs = self.model(input_data)
File "/home/pengfang/.conda/envs/svtas/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "/home/pengfang/SVTAS-svtas-paper/model/architectures/segmentation/transeger.py", line 77, in forward
text_output = self.text_backbone(text_input)
File "/home/pengfang/.conda/envs/svtas/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "/home/pengfang/SVTAS-svtas-paper/model/architectures/general/encoder_2_decoder.py", line 64, in forward
x = self.encoder(x, masks)
File "/home/pengfang/.conda/envs/svtas/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "/home/pengfang/SVTAS-svtas-paper/model/backbones/language/learner_prompt.py", line 87, in forward
print('labels',labels.size())
AttributeError: 'NoneType' object has no attribute 'size'

在运行尝试运行transeger时，发现text_backbone的输入无法初始化，因为是第一个片段，所以class Transger中的self.last_clip_labels为None，无法进行文本编码，请问是代码的bug还是我的输入需要调整？