V1.1.0

This release coincides with the release of the 5 2021 challenge leaderboards for EPIC-KITCHENS-100.

Major changes compared to V1.0.0:

• Addition of Unsupervised Domain Adaptation validation files (ad808df)
• Addition of sentence files for the Multi-Instance Retrieval Challenge (f36021a)
• Adds missing metadata for returning/new kitchens (65192d4)
• Adds information on tail classes and unseen participant IDs (be2e688)
• Adds video information for fps, duration, resolution (6c92b84)

See the readme.md for more information about the new files.

The version of the python pickles can be checked as follows to ensure you have the correct version:

>>> import pandas as pd
>>> train = pd.read_pickle('EPIC_100_train.pkl')
>>> train._metadata['version_number']
'v1.1.0'

There is also a commit number (accessible via train._metadata['commit_hash']) should you need to check specifically which commit these pickles were generated from.

V1.0.0

V1.0.0

First release of the annotations for the EPIC-KITCHENS-100 dataset.

This includes annotation files for the three main splits:

• EPIC_100_train.csv for the training set.
• EPIC_100_validation.csv for the validation set.
• EPIC_100_test_timestamps.csv for the testing set.

Note that all narrations in any of the test sets do not have labelled verb/noun classes for the challenges.

As well as the verb/noun class files:

• EPIC_100_verb_classes.csv for the verb classes and categories.
• EPIC_100_noun_classes.csv for the noun classes and categories.

The annotations for the Unsupervised Domain Adaptation (UDA) and Action retrieval can be respectively found under UDA_annotations and retrieval_annotations.

See the readme.md for more information about the annotations.

We also provide pickle files for the annotations csvs. The version of the python pickles can be checked as follows:

>>> import pandas as pd
>>> train = pd.read_pickle('EPIC_100_train.pkl')
>>> train._metadata['version_number']
'v1.0.0'

There is also a commit number (accessible via train._metadata['commit_hash']) should you need to check specifically which commit these pickles were generated from.