search

cmhungsteve / TA3N

[ICCV 2019 (Oral)] Temporal Attentive Alignment for Large-Scale Video Domain Adaptation (PyTorch)
https://arxiv.org/abs/1907.12743
MIT License
259 stars 41 forks source link

Code for obtaining the baseline results. #23

Closed AadSah closed 3 years ago

AadSah commented 3 years ago

Hi @cmhungsteve

Thanks for the great work. I wanted to execute and obtain the baseline results mentioned in the paper (DANN + [TempPooling/TempRelation]), can you please let me know if you can make the code public?

Thanks

cmhungsteve commented 3 years ago

DANN + TemPooling: frame_aggregation=avgpool use_target=uSv adv_DA=RevGrad

DANN + TemRelation: frame_aggregation=trn-m use_target=uSv adv_DA=RevGrad

and set other XXX_DA as none.

AadSah commented 3 years ago

Thanks a lot @cmhungsteve It would be great if you could also share the arguments for obtaining the 'source-only' and 'target-only' results. :)

cmhungsteve commented 3 years ago

Just set use_target=none, and then modify dataset_source. For example, if you are working on hmdb_ucf and want to train and test on ucf101 only, then set dataset_source=ucf101, otherwise set dataset_source=hmdb51. In another word, this all depends on how you define your source and target datasets.

avijit9 commented 3 years ago

Do I need to set use_attn to none when reproducing DANN results from 3, 4, 5?

cmhungsteve commented 3 years ago

Yes