AlphaAction code is not updated

[netanellavisdris]

Hi, When trying to run the repository, I've got this error :

Traceback (most recent call last):
  File "train_net.py", line 12, in <module>
    from alphaction.dataset import make_data_loader
  File "/home/netanell/work/STMixer/alphaction/dataset/__init__.py", line 1, in <module>
    from .build import make_data_loader
  File "/home/netanell/work/STMixer/alphaction/dataset/build.py", line 5, in <module>
    from . import datasets as D
  File "/home/netanell/work/STMixer/alphaction/dataset/datasets/__init__.py", line 3, in <module>
    from .ava_kinetics import AvaKinetics
ModuleNotFoundError: No module named 'alphaction.dataset.datasets.ava_kinetics'

It seems that there is mismatch in alphaction code used here.

Best, Netanell

[atao99]

Thanks for pointing out this issue. We've fixed it.

[netanellavisdris]

Thanks for the quick fix.

Results are close to the reported one with pretrained weights We got 0.3489 mAP@IOU=0.5 on AVA eval with VMAE-Vit-BV2 While in the paper you report 36.1 mAP. Does the published weights are the one you got AVA v2.2 results?

Thanks, Netanell.

[atao99]

We did the evaluation again and we found no problems here. Could you confirm whether it is only the model with VMAEv2-ViT-B backbone that has the problem or other models also have the problem?

[netanellavisdris]

We only check VMAEv2-ViT-B backbone and got these results

2023-05-11 19:37:36,494 alphaction.inference INFO: { 'PascalBoxes_PerformanceByCategory/[email protected]/answer phone': 0.819466937633075,
  'PascalBoxes_PerformanceByCategory/[email protected]/bend/bow (at the waist)': 0.5102316213722198,
  'PascalBoxes_PerformanceByCategory/[email protected]/carry/hold (an object)': 0.6544299296389664,
  'PascalBoxes_PerformanceByCategory/[email protected]/climb (e.g., a mountain)': 0.11858355788898292,
  'PascalBoxes_PerformanceByCategory/[email protected]/close (e.g., a door, a box)': 0.24730749081212594,
  'PascalBoxes_PerformanceByCategory/[email protected]/crouch/kneel': 0.37864294304590784,
  'PascalBoxes_PerformanceByCategory/[email protected]/cut': 0.23370729929645434,
  'PascalBoxes_PerformanceByCategory/[email protected]/dance': 0.6692521126811624,
  'PascalBoxes_PerformanceByCategory/[email protected]/dress/put on clothing': 0.08350067878855291,
  'PascalBoxes_PerformanceByCategory/[email protected]/drink': 0.378015286264672,
  'PascalBoxes_PerformanceByCategory/[email protected]/drive (e.g., a car, a truck)': 0.632076390189445,
  'PascalBoxes_PerformanceByCategory/[email protected]/eat': 0.4403115326469822,
  'PascalBoxes_PerformanceByCategory/[email protected]/enter': 0.0489432489651446,
  'PascalBoxes_PerformanceByCategory/[email protected]/fall down': 0.1958661876487952,
  'PascalBoxes_PerformanceByCategory/[email protected]/fight/hit (a person)': 0.5417155907456872,
  'PascalBoxes_PerformanceByCategory/[email protected]/get up': 0.40821602451137107,
  'PascalBoxes_PerformanceByCategory/[email protected]/give/serve (an object) to (a person)': 0.16735980046346485,
  'PascalBoxes_PerformanceByCategory/[email protected]/grab (a person)': 0.12585601545399266,
  'PascalBoxes_PerformanceByCategory/[email protected]/hand clap': 0.4121450206697173,
  'PascalBoxes_PerformanceByCategory/[email protected]/hand shake': 0.14808346164749678,
  'PascalBoxes_PerformanceByCategory/[email protected]/hand wave': 0.1351064377829511,
  'PascalBoxes_PerformanceByCategory/[email protected]/hit (an object)': 0.1565221968532169,
  'PascalBoxes_PerformanceByCategory/[email protected]/hug (a person)': 0.3131640025620841,
  'PascalBoxes_PerformanceByCategory/[email protected]/jump/leap': 0.19169379800173414,
  'PascalBoxes_PerformanceByCategory/[email protected]/kiss (a person)': 0.4511101048191475,
  'PascalBoxes_PerformanceByCategory/[email protected]/lie/sleep': 0.5070461124074798,
  'PascalBoxes_PerformanceByCategory/[email protected]/lift (a person)': 0.16041297378061492,
  'PascalBoxes_PerformanceByCategory/[email protected]/lift/pick up': 0.046137077364503586,
  'PascalBoxes_PerformanceByCategory/[email protected]/listen (e.g., to music)': 0.05064308377278025,
  'PascalBoxes_PerformanceByCategory/[email protected]/listen to (a person)': 0.6930314054518483,
  'PascalBoxes_PerformanceByCategory/[email protected]/martial art': 0.4792668449573899,
  'PascalBoxes_PerformanceByCategory/[email protected]/open (e.g., a window, a car door)': 0.33303843175451436,
  'PascalBoxes_PerformanceByCategory/[email protected]/play musical instrument': 0.6854144356424285,
  'PascalBoxes_PerformanceByCategory/[email protected]/point to (an object)': 0.008190591507090613,
  'PascalBoxes_PerformanceByCategory/[email protected]/pull (an object)': 0.030260108898131854,
  'PascalBoxes_PerformanceByCategory/[email protected]/push (an object)': 0.08657010175807288,
  'PascalBoxes_PerformanceByCategory/[email protected]/push (another person)': 0.049630232552342675,
  'PascalBoxes_PerformanceByCategory/[email protected]/put down': 0.04525505516075375,
  'PascalBoxes_PerformanceByCategory/[email protected]/read': 0.46418561337472286,
  'PascalBoxes_PerformanceByCategory/[email protected]/ride (e.g., a bike, a car, a horse)': 0.4866691088045868,
  'PascalBoxes_PerformanceByCategory/[email protected]/run/jog': 0.609256187694392,
  'PascalBoxes_PerformanceByCategory/[email protected]/sail boat': 0.2834965157071825,
  'PascalBoxes_PerformanceByCategory/[email protected]/shoot': 0.4432708318403956,
  'PascalBoxes_PerformanceByCategory/[email protected]/sing to (e.g., self, a person, a group)': 0.4226643026272972,
  'PascalBoxes_PerformanceByCategory/[email protected]/sit': 0.8004915834582601,
  'PascalBoxes_PerformanceByCategory/[email protected]/smoke': 0.3009767764066721,
  'PascalBoxes_PerformanceByCategory/[email protected]/stand': 0.8444927323891764,
  'PascalBoxes_PerformanceByCategory/[email protected]/swim': 0.7150849026538953,
  'PascalBoxes_PerformanceByCategory/[email protected]/take (an object) from (a person)': 0.0909614218087088,
  'PascalBoxes_PerformanceByCategory/[email protected]/take a photo': 0.05589583957630578,
  'PascalBoxes_PerformanceByCategory/[email protected]/talk to (e.g., self, a person, a group)': 0.8221556167362185,
  'PascalBoxes_PerformanceByCategory/[email protected]/text on/look at a cellphone': 0.2101792269015712,
  'PascalBoxes_PerformanceByCategory/[email protected]/throw': 0.0739420237625154,
  'PascalBoxes_PerformanceByCategory/[email protected]/touch (an object)': 0.3704596901246441,
  'PascalBoxes_PerformanceByCategory/[email protected]/turn (e.g., a screwdriver)': 0.04368915509960961,
  'PascalBoxes_PerformanceByCategory/[email protected]/walk': 0.78746549181684,
  'PascalBoxes_PerformanceByCategory/[email protected]/watch (a person)': 0.7261317750622113,
  'PascalBoxes_PerformanceByCategory/[email protected]/watch (e.g., TV)': 0.19887543058009893,
  'PascalBoxes_PerformanceByCategory/[email protected]/work on a computer': 0.358959317415066,
  'PascalBoxes_PerformanceByCategory/[email protected]/write': 0.1895039611343681,
  'PascalBoxes_Precision/[email protected]': 0.3489168605061007}

Will continue (it will take time) to other backbones.