mmaction

mmaction copied to clipboard

Thank you for your good tools! Could you provide a technical report for more detailed understanding?

zhangyuan1994511

MMAction Roadmap

1

## Tasks - [x] Video classificaiton * [x] 2D CNN (RGB, Two-stream CNN) * [x] 3D CNN (I3D) * [x] TSN sampling - [x] Temporal Action Detection * [x] SSN...

yjxiong

i use cpu only,but cannot compile?how to resolve it？

1

_build_ext.run(self) File "/opt/conda/lib/python3.7/distutils/command/build_ext.py", line 340, in run self.build_extensions() File "/opt/conda/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 353, in build_extensions build_ext.build_extensions(self) File "/opt/conda/lib/python3.7/distutils/command/build_ext.py", line 449, in build_extensions self._build_extensions_serial() File "/opt/conda/lib/python3.7/distutils/command/build_ext.py", line 474, in _build_extensions_serial self.build_extension(ext) File...

dotsonliu

Thanks for providing this comprehensive code base. When using a single machine with 8 GPUs to train an I3D model, the program will hang at the end of an epoch,...

gfjiangly

errors : all tensors must be on devices[0]

17

Hi. I have two titan xp on my single machine. i run the command as below. all other settings as github provided without any changes: **./tools/dist_test_recognizer.sh configs/ucf101/tsn_rgb_bninception.py ./modelzoo/tsn_2d_rgb_bninception_seg3_f1s1_b32_g8-98160339.pth 2** result:...

sijun-zhou

Temporal Action Localization for THMOS'14

1

Hi! I am trying to carry out temporal action localization for THUMOS14 dataset. I already have RGB plus OF frames using dense flow. I am using pre-trained model given for...

pranavgundewar

inference one video

1

can i inference my own video and get the result?

juvien

Hi, I am trying to train thumos-14 dataset using optical-flow. Using the given config, I can get it working for RGB modality. However, I get this following error when I...

manmadan03

Training with a single GPU

1

Heads up that validate doesn't do anything in `train_recognizer.py` if it's not distributed training. The keyword arg isn't used in the function and trying to use it in the config...

cinjon

File: `data_tools/build_rawframes.py:140` Issue: `--resume` flag redoes all the completed videos. Reason: `fullpath_list` has elements of the form `.../src_dir/v.mp4`, while `done_fullpath_list` has elements of the form `.../out_dir/v`. They never have anything...

arpanmangal