xlliu7 / tadtr Goto Github PK

Hi! I'm really interested in using this work for action detection - is there any way I could get access to your training scripts and pretrained weights?

Hi, first thanks for your great work.
I am trying to reproduce your results in ActivityNet. I follow the operations in your paper. Using TSP features and add some codes in Dataset module. I can run through whole process in ActivityNet but i just cannot get results as good as you present in the paper. For me, the results drop all about 3-4%.
I am wondering whether you have planning to open source the train code for ActivityNet?

Could you please share the counting code? Thanks!

Thanks for your great work. When will the training & inference code release? Can you give an approximate date? Thanks!

I noticed that there are mismatched keys name in weight_dict, effectively making the losses calculation skipped loss_segments and loss_actionness in this line:

Looking at the weight_dict, loss_seg is used rather than loss_segments

For actionness, it's assigned as loss_iou instead of loss_actionness, which replaced the loss_iou by segments loss.

Are these bugs? Could you confirm it? Thanks.

Hi, xiaolong. I'm very interested in your work. As you mentioned in another issue, you use the I3D features form P-GCN for the Thumos14 experiment. I find that some features for the same video have different sizes so that I can't concat them directly. And the diff is always 1. Have you ever met this situation. If ever, how you deal with it? Thx~

https://github.com/xlliu7/TadTR/blob/master/models/transformer.py#L287
这里是在decoder的forward里面计算reference point，
但是在这里
https://github.com/xlliu7/TadTR/blob/master/models/tadtr.py#L185
又重新计算了一次，这两部分应该是重复了

The claimed improvement from actionness regression does not seem to materialize based on my implementation using this code repository. The results with and without actionness regression are very similar.

Upon inspecting the implementation, I noticed a potential issue:

Even after correcting this issue, there was still no performance improvement from the actionness regression in my runs (the performance drops a lot actually). Upon my inspection, that because the actionness regression suffers from a serious label imbalance problem as most target IoUs are zero.

请问大佬源码什么时候公布呢？

Hello, thank you for your good work!
I want to know how to generate th14_i3d2s_ft_info.json for thumos14 video features. And how to compute ''feature_length", "feature_second" and "feature_fps" for each video?

Hello, can you provide the TSN features after linear interpolation of activitynet1.3？

when i train on my dataset ,something wrong happened.the map_raw was nan
can you give me some suggestions?
thanks!

Hi, thank you for your brilliant work!

Can you provided your test results on ActivityNet1.3 dataset?

Thank you very much!

Hello,Thank u for your work!
I want to know how feature_length can be read directly from the video feature files，because I use my own dataset to try this code.

Hi @xlliu7,

Interesting paper! I want to know how to combine your model with the classifier? e.g. PGCN in Table 1.
Would you mind sharing the code? Thanks.

Hi Could you tell me how to get the I3D 2stream Feature ? Thanks

作者您好！最近准备研究这个方向，请问有没有在单个视频上进行推理的代码？

Can you provide the link to download the features.

I'm trying to train on relatively small datasets, mix-up is one way to reduce it from overfitting, but it seems like focal loss is not designed to works with label with probabilities. It seems that this line

Do you have any idea how to modify focal loss for label with probabilities?

Dear researchers,

Thank you for your work.

The links for your networks weights don t work, which prevent us to reproduce your work.

best regards,

Thanks open source for this good work.

But, I met a problem.

models/ops/temporal_deform_attn/functions/temporal_deform_attn_func.py", line 40, in backward value, value_spatial_shapes, value_level_start_index, sampling_locations, attention_weights, grad_output, ctx.seq2col_step) RuntimeError: Not implemented

I wonder if it is convenient for you to answer.

Do you have a planed time to release the code of E2E-TAD?

Hello,

thank you for sharing the code. I checked the code and all of the seeds are set.
I further added torch.backends.cudnn.deterministic = True and torch.backends.cudnn.benchmark = False to make the code produce the same results in different runs. However, the results differ between different runs.
Do you have any idea why?

Thanks in advance.

Dear @xlliu7, thank you for your valuable contribution to the community.
I know cleaning code and supporting all datasets require much work.
However, I would greatly appreciate it if you were to release the rest of the code for reproducing results in HACS and ActivityNet.

Could you kindly let me know the time horizon for this?

Best,
Mattia

I encountered the above error when trying to run the demo code.

Hi. The arxiv paper says code will be made available here. Do you have a date when you plan to release it? Thanks.

Hello, thanks for the nice work.
I have sent an email to request you the codes for ActivityNet.
Could you share your codes?

Best wishes.