search

Alvin-Zeng / PGCN

Graph Convolutional Networks for Temporal Action Localization (ICCV2019)
319 stars 66 forks source link

How to run this model on a New datasets #41

Closed makecent closed 3 years ago

makecent commented 3 years ago

I would like to test your model on a new TAL dataset collected by our laboratory. Hence we want to know how should we prepare the dataset directory and ground truth files. Any suggestions will be very helpful!

rahulrajus commented 3 years ago

Hey! Have you made any progress on training on new datasets?

makecent commented 3 years ago

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

menghuaa commented 3 years ago

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! How you generate the proposals?

makecent commented 3 years ago

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! How you generate the proposals?

I used the BMN, which is supported in the mmaction2 repository.

menghuaa commented 3 years ago

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! How you generate the proposals?

I used the BMN, which is supported in the mmaction2 repository.

Hi!How you generate the features of proposals?

makecent commented 3 years ago

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! Have you made any progress on training on new datasets?

I made it. I use the mmaction2 repository to generate the proposals for the new dataset and use them to train the PGCN successfully. PGCN works pretty well.

Hey! How you generate the proposals?

I used the BMN, which is supported in the mmaction2 repository.

Hi!How you generate the features of proposals?

Just following the PGCN: using a I3D pre-trained on Kinetics to extract features on the new dataset. You may refer to step 6 here.

try-harder12 commented 8 months ago

Hello, I would like to ask how you processed the proposal result generated by BMN to convert it into the proposal format required by PGCN？ @makecent Looking forward to your help！

henryrobbins commented 6 months ago

@try-harder12 did you have any success with this? I'm about to attempt the same thing myself. In the BSN README, it mentions a script tools/data/activitynet/convert_proposal_format.py which can convert into the proposal format for SSN or P-GCN.