Home

Awesome

EPIC-KITCHENS Action Recognition baselines

Train/Val/Test splits and annotations are available at Annotations Repo

To participate and submit to this challenge, register at Action Recognition Codalab Challenge

Released Models

Result data formats

We support two formats for model results.

Either of these formats can saved via torch.save with .pt suffix or with pickle.dump with a .pkl suffix.

We provide two example files (from a TSN trained on RGB) following both these formats for reference.

Note that either of these layouts can be stored in a .pkl/.pt file--the dict format doesn't necessarily have to be in a .pkl.

Evaluating model results

We provide an evaluation script to compute the metrics we report in the paper on the validation set. You will also need to clone the annotations repo

$ python evaluate.py \
    results.pt \
    /path/to/annotations/EPIC_100_validation.pkl \
    --tail-verb-classes-csv /path/to/annotations/EPIC_100_tail_verbs.csv \
    --tail-noun-classes-csv /path/to/annotations/EPIC_100_tail_nouns.csv \
    --unseen-participant-ids /path/to/annotations/EPIC_100_unseen_participant_ids_test.csv \

all_action_accuracy_at_1: ...
all_action_accuracy_at_5: ...
all_noun_accuracy_at_1: ...
all_noun_accuracy_at_5: ...
all_verb_accuracy_at_1: ...
all_verb_accuracy_at_5: ...
tail_action_accuracy_at_1: ...
tail_noun_accuracy_at_1: ...
tail_verb_accuracy_at_1: ...
unseen_action_accuracy_at_1: ...
unseen_noun_accuracy_at_1: ...
unseen_verb_accuracy_at_1: ...

Creating competition submission files

Once you have evaluated your model on the test set, you can simply run this script to generate a zip file containing the JSON representation of your model's scores that can be submitted to the challenge leaderboard.

$ python create_submission.py test_results.pt test_results.zip

An example of the output this script produces: https://www.dropbox.com/s/sxxgyscziryeno1/tsn_fused_split%3Dtest.zip?dl=1 This is the submission file for the modality fused results of TSN model.

License

Copyright University of Bristol. The repository is published under the Creative Commons Attribution-NonCommercial 4.0 International License. This means that you must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. You may not use the material for commercial purposes.