Home

Awesome

EPIC KITCHENS-100 Dataset

<!-- start badges -->

CircleCI GitHub release arXiv-2006.13256

<!-- end badges -->

EPIC-KITCHENS-100 is the largest dataset in first-person (egocentric) vision; itself an extension of the EPIC-KITCHENS-55 dataset (formally known as EPIC-KITCHENS-2018).

Authors

Dima Damen (1) Hazel Doughty (1) Giovanni Maria Farinella (2) Antonino Furnari (2) Evangelos Kazakos (1) Jian Ma (1) Davide Moltisanti (1) Jonathan Munro (1) Toby Perrett (1) Will Price (1) Michael Wray (1)

Contact: uob-epic-kitchens@bristol.ac.uk

Citing

When using the dataset, kindly reference:

@article{damen2022rescaling,
   title={Rescaling Egocentric Vision},
   author={Damen, Dima and Doughty, Hazel and Farinella, Giovanni Maria  and and Furnari, Antonino 
           and Ma, Jian and Kazakos, Evangelos and Moltisanti, Davide and Munro, Jonathan 
           and Perrett, Toby and Price, Will and Wray, Michael},
           journal={International Journal of Computer Vision},
           volume={130},
           number={1},
           pages={33--55},
           year={2022},
           publisher={Springer}
} 

Erratum

Important: We have recently detected an error in our pre-extracted RGB and Optical flow frames for two videos in our dataset. This does not affect the videos themselves or any of the annotations in this github. However, if you've been using our pre-extracted frames, we below detail how you can fix the error at your end, until we publish replacement frames for downloading.

Download the videos P01_109.MP4 and P27_103.MP4. Then set up a directory like so:

$ mkdir -p rgb/{P01_109,P27_103}
$ mkdir -p flow/{P01_109,P27_103}
$ mkdir videos
$ mv /path/to/{P01_109,P27_103}.MP4 videos

You will need docker setup on your machine to extract the frames and flow.

RGB

$ docker run --gpus "device=0" \
     -it \
     --rm \
     -v "$PWD:/workspace" \
     willprice/nvidia-ffmpeg \
     -hwaccel cuvid \
     -c:v hevc_cuvid \
     -i /workspace/videos/P27_103.MP4 \
     -vf 'scale_npp=-2:256:interp_algo=super,hwdownload,format=nv12' \
     -qscale:v 4 \
     -r 50 /workspace/rgb/P27_103/frame_%010d.jpg

$ docker run --gpus "device=0" \
     -it \
     --rm \
     -v "$PWD:/workspace" \
     willprice/nvidia-ffmpeg \
     -hwaccel cuvid \
     -c:v hevc_cuvid \
     -i /workspace/videos/P01_109.MP4 \
     -vf 'scale_npp=-2:256:interp_algo=super,hwdownload,format=nv12' \
     -qscale:v 4 \
     -r 50 /workspace/rgb/P01_109/frame_%010d.jpg

Flow

$ docker run --gpus "device=0" \
     -it \
     --rm \
     -v "$PWD/rgb/P01_109:/input" \
     -v "$PWD/flow/P01_109:/output" \
     willprice/furnari-flow \
     frame_%010d.jpg -g 0 -s 1 -d 1 -b 8

$ docker run --gpus "device=0" \
     -it \
     --rm \
     -v "$PWD/rgb/P27_103:/input" \
     -v "$PWD/flow/P27_103:/output" \
     willprice/furnari-flow \
     frame_%010d.jpg -g 0 -s 1 -d 1 -b 8

Index

Dataset Details

The EPIC-KITCHENS-100 dataset is an extension of the EPIC-KITCHENS-55 dataset. Videos are distinguished as follows:

The dataset currently has 6 active benchmarks:

We provide csv files for the train/val/test sets of each benchmark detailed below for ease of use, see Important Files for more information.

Ground truth is provided for action segments as action/verb/noun labels along with the start and end times of the segment.

We also provide automatic annotations in the form of object masks and hand/object BBoxes. See automatic annotations for more details.

back to top

Quick Start

Here you can download the annotation files for all of the challenges. For more information on each challenge, please see the paper here. A download script is provided for the videos, RGB Frames and Flow frames here.

Action Recognition Challenge

  1. Download the videos/RGB/Flow frames here with the following command:
python epic_downloader.py --videos --rgb-frames --flow-frames
  1. Download the Action Recognition train/val/test files.
  2. Enjoy the EPIC-KITCHENS-100 dataset in your favourite action recognition model, see the paper for details on the models we used for this baseline. Models trained on EPIC-KITCHENS-55 can be found here as a starting point.

Weakly Supervised Action Recognition Challenge

  1. Download the videos/RGB/Flow frames here with the following command:
python epic_downloader.py --videos --rgb-frames --flow-frames
  1. This challenge uses the Action Recognition files, download the train/val/test files.
  2. The weakly supervised challenge uses the narration timestamp, not the the start/end times of the action. Therefore a simple baseline would be to modify an action recognition model to use the surrounding 5s worth of frames. See the paper for details on the models we used for this baseline.

Action Detection Challenge

  1. Download the videos/RGB/Flow frames here with the following command:
python epic_downloader.py --videos --rgb-frames --flow-frames
  1. This challenge uses the Action Recognition files, download the train/val/test files.
  2. Train an action proposal network on the EPIC-KITCHENS-100 train set, for example this model. This model predicts action-agnostic segments which still need to be classified.
  3. Use your favourite action recognition model to classify the proposals (example models).

Action Anticipation Challenge

  1. Download the videos/RGB/Flow frames here with the following command:
python epic_downloader.py --videos --rgb-frames --flow-frames
  1. This challenge uses the Action Recognition files, download the train/val/test files.
  2. A simple baseline for this task is to train an action recognition model (example models here) on the 5 seconds that precede an action with a 1 second gap. For example, an action that starts at 20.00s in a video would see frames between 14.00s and 19.00s.

Unsupervised Domain Adaptation Challenge

The unsupervised domain adaptation challenge tests how models can cope with similar data collected 2 years later on the task of action recognition.

  1. Download the videos/RGB/Flow frames here with the following command:
python epic_downloader.py --videos --rgb-frames --flow-frames --domain-adaptation
  1. Download the Unsupervised Domain Adaptation source train/target train/source_test/target test/source val/target val files.
  2. Extract video features (for all six splits) using an off-the-shelf model trained on EPIC-KITCHENS-55 (example model).
  3. A simple baseline is using a domain discriminator (prediciting whether a video came from the source, EPIC-KITCHENS-55, or the target, EPIC-KITCHENS-100) to align the two domains. See the paper for details on the models we used for this baseline.

IMPORTANT NOTE ON HYPER-PARAMETER TUNING. As the target domain is unlabelled, the training splits cannot be used for hyper-parameter tuning. You must use the validation splits to choose hyper-parameters. The procedure for hyper-parameter tuning and training is as follows:

  1. Train your model on source val with unlabelled data from target val.
  2. Evaluate your model on target val using the labels provided (these labels should not be used during training).
  3. Select hyper-parameters based on the performance on target val.
  4. Re-train your model on source train/target train with selected hyper-parameters.
  5. Evaluate the re-trained model on target test to produce action predictions for the challenge leaderboard.

It is optional but highly ecouraged to evalute the performance on source_test to compare source domain performances.

Multi-Instance Retrieval Challenge

NOTE 30/09/2020 There was an error in the creation of the sentence files for the retrieval challenge. Please download the new sentence dataframes.

  1. Download the videos/RGB/Flow frames here with the following command:
python epic_downloader.py --videos --rgb-frames --flow-frames --action-retrieval
  1. Download the Multi-Instance Retrieval train/test files.
  2. Extract video features (for both the train and test set) using an off-the-shelf model trained on EPIC-KITCHENS-55 (example model).
  3. Extract word2vec features for the captions from both the train and test set (example models).
  4. Enjoy the EPIC-KITCHENS-100 dataset in your favourite video retrieval model, see the paper for details on the models we used for this baseline.

back to top

Important Files

For ease of use, download scripts are provided to download the videos and RGB/Flow frames. (see file downloads for more details). We direct the reader to RDSF for the full release of videos and RGB/Flow frames. We provide html and pdf alternatives to this README which are auto-generated.

Additional Files

back to top

File Structure

EPIC_100_train.csv

This CSV file contains the action annotations for the training set and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP01_01_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_01ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:01.089Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.14Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:03.37End time in HH:mm:ss.SS of the action segment.
start_frameint8Start frame of the action.
stop_frameint202End frame of the action.
narrationstringopen doorTranscribed description of the English narration provided by the participant.
verbstringopenParsed verb from the narration.
verb_classint3Numeric ID of the verb's class.
nounstringdoorFirst parsed noun from the narration.
noun_classint3Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[door]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[3]Numeric ID of all of the parsed noun's classes.

Back to Important Files

EPIC_100_validation.csv

This CSV file contains the action annotations for the validation set and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP01_01_11Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_11ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:00.560Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.00Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:01.89End time in HH:mm:ss.SS of the action segment.
start_frameint1Start frame of the action.
stop_frameint113End frame of the action.
narrationstringtake plateTranscribed description of the English narration provided by the participant.
verbstringtakeParsed verb from the narration.
verb_classint0Numeric ID of the verb's class.
nounstringplateFirst parsed noun from the narration.
noun_classint2Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[plate]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[2]Numeric ID of all of the parsed noun's classes.

Back to Important Files

EPIC_100_test_timestamps.csv

This CSV file contains the action annotations for the testing set and contains 9 columns:

Column NameTypeExampleDescription
narration_idstringP01_101_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_101ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:02.851Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:02.86Start time in HH:mm:ss.SSS of the action segment.
stop_timestampstring00:00:03.87End time in HH:mm:ss.SSS of the action segment.
start_frameint143Start frame of the action.
stop_frameint193End frame of the action.

Back to Important Files

EPIC_100_noun_classes.csv

This CSV file contains information on the 300 noun classes and contains 4 columns.

Column NameTypeExampleDescription
idint222Unique ID for the noun class.
keystringlabelKey used for the noun class (all keys are a member of their own class).
instanceslist of string (1 or more)"['label', 'sticker']"All nouns within the class, including the key.
categorystringmaterialsName of the higher-level noun category that this noun class belongs to.

Back to Important Files

EPIC_100_verb_classes.csv

This CSV file contains information on the 97 verb classes and contains 4 columns.

Column NameTypeExampleDescription
idint79Unique ID for the verb class.
keystringlet-goKey used for the verb class (all keys are a member of their own class).
instanceslist of string (1 or more)"['let', 'let-go']"All verbs within the class, including the key.
categorystringleaveName of the higher-level verb category that this verb class belongs to.

Back to Important Files

EPIC_100_uda_source_train.csv

This CSV file contains the action annotations for the source training set used for Unsupervised Domain Adaptation and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP01_01_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_01ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:01.089Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.14Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:03.37End time in HH:mm:ss.SS of the action segment.
start_frameint8Start frame of the action.
stop_frameint202End frame of the action.
narrationstringopen doorTranscribed description of the English narration provided by the participant.
verbstringopenParsed verb from the narration.
verb_classint3Numeric ID of the verb's class.
nounstringdoorFirst parsed noun from the narration.
noun_classint3Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[door]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[3]Numeric ID of all of the parsed noun's classes.

Note that this file contains only videos from EPIC-KITCHENS-55 which is used as the source domain.

See here for more details on the unsupervised domain adaptation challenge.

Back to Important Files

EPIC_100_uda_source_test_timestamps.csv

This CSV file contains the action annotations for the source testing set used for Unsupervised Domain Adaptation and contains 9 columns:

Column NameTypeExampleDescription
narration_idstringP01_11_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_11ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:00.560Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.00Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:01.89End time in HH:mm:ss.SS of the action segment.
start_frameint1Start frame of the action.
stop_frameint113End frame of the action.

Note that this file contains only videos from EPIC-KITCHENS-55 which is used as the source domain.

See here for more details on the unsupervised domain adaptation challenge.

Back to Important Files

EPIC_100_uda_target_train_timestamps.csv

This CSV file contains the action annotations for the target training set used for Unsupervised Domain Adaptation and contains 9 columns:

Column NameTypeExampleDescription
narration_idstringP01_102_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_102ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:01.100Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.54Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:02.23End time in HH:mm:ss.SS of the action segment.
start_frameint27Start frame of the action.
stop_frameint111End frame of the action.

Note that this file contains only videos from EPIC-KITCHENS-100 which is used as the target domain.

See here for more details on the unsupervised domain adaptation challenge.

Back to Important Files

EPIC_100_uda_target_test_timestamps.csv

This CSV file contains the action annotations for the target testing set used for Unsupervised Domain Adaptation and contains 9 columns:

Column NameTypeExampleDescription
narration_idstringP01_101_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_101ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:02.851Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:02.86Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:03.87End time in HH:mm:ss.SS of the action segment.
start_frameint143Start frame of the action.
stop_frameint193End frame of the action.

Note that this file contains only videos from EPIC-KITCHENS-100 which is used as the target domain.

See here for more details on the unsupervised domain adaptation challenge.

Back to Important Files

EPIC_100_uda_source_val.csv

This CSV file contains the action annotations for the source validation set used for Unsupervised Domain Adaptation and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP03_02_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP03ID of the participant (unique per participant).
video_idstringP03_02ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:04.310Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:03.29Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:04.26End time in HH:mm:ss.SS of the action segment.
start_frameint197Start frame of the action.
stop_frameint255End frame of the action.
narrationstringput lunch boxTranscribed description of the English narration provided by the participant.
verbstringputParsed verb from the narration.
verb_classint1Numeric ID of the verb's class.
nounstringbox:lunchFirst parsed noun from the narration.
noun_classint23Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[box:lunch]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[23]Numeric ID of all of the parsed noun's classes.

Note that this file contains only videos from EPIC-KITCHENS-55 which is used as the source domain for validation.

See here for more details on the unsupervised domain adaptation challenge.

Back to Important Files

EPIC_100_uda_target_val.csv

This CSV file contains the action annotations for the target validation set used for Unsupervised Domain Adaptation and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP03_101_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP03ID of the participant (unique per participant).
video_idstringP03_101ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:02.877Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:02.60Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:03.86End time in HH:mm:ss.SS of the action segment.
start_frameint130Start frame of the action.
stop_frameint193End frame of the action.
narrationstringturn on tapTranscribed description of the English narration provided by the participant.
verbstringturn-onParsed verb from the narration.
verb_classint6Numeric ID of the verb's class.
nounstringtapFirst parsed noun from the narration.
noun_classint0Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[tap]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[23]Numeric ID of all of the parsed noun's classes.

Note that this file contains only videos from EPIC-KITCHENS-100 which is used as the target domain for validation.

See here for more details on the unsupervised domain adaptation challenge.

Back to Important Files

EPIC_100_retrieval_train.csv

This CSV file contains the action annotations for the action retrieval training set and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP01_01_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_01ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:01.089Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.14Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:03.37End time in HH:mm:ss.SS of the action segment.
start_frameint8Start frame of the action.
stop_frameint202End frame of the action.
narrationstringopen doorTranscribed description of the English narration provided by the participant.
verbstringopenParsed verb from the narration.
verb_classint3Numeric ID of the verb's class.
nounstringdoorFirst parsed noun from the narration.
noun_classint3Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[door]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[3]Numeric ID of all of the parsed noun's classes.

Back to Important Files

EPIC_100_retrieval_train_sentence.csv

This CSV file contains the caption annotations for the action retrieval training set and contains 6 columns:

Column NameTypeExampleDescription
narration_idstringP01_01_0Unique ID for the caption (corresponding to the original action).
narrationstringopen doorTranscribed description of the English narration provided by the participant.
verb_classint3Numeric ID of the verb's class.
noun_classeslist of int (1 or more)[3]Numeric ID of the all noun classes in the narration.
verbstringopenParsed verb from the narration.
nounslist of string (1 or more)[door]All parsed nouns in the narration.

Back to Important Files

EPIC_100_retrieval_test.csv

This CSV file contains the action annotations for the action retrieval testing set and contains 15 columns:

Column NameTypeExampleDescription
narration_idstringP01_11_0Unique ID for the segment as a string with participant ID and video ID.
participant_idintP01ID of the participant (unique per participant).
video_idstringP01_11ID of the video where the segment originated from (unique per video).
narration_timestampstring00:00:00.560Timestamp of when the original narration was recorded in HH:mm:ss.SSS.
start_timestampstring00:00:00.00Start time in HH:mm:ss.SS of the action segment.
stop_timestampstring00:00:01.89End time in HH:mm:ss.SS of the action segment.
start_frameint8Start frame of the action.
stop_frameint113End frame of the action.
narrationstringtake plateTranscribed description of the English narration provided by the participant.
verbstringtakeParsed verb from the narration.
verb_classint0Numeric ID of the verb's class.
nounstringplateFirst parsed noun from the narration.
noun_classint2Numeric ID of the first noun's class.
all_nounslist of string (1 or more)[plate]List of all parsed nouns within the narration.
all_noun_classeslist of int (1 or more)[2]Numeric ID of all of the parsed noun's classes.

Back to Important Files

EPIC_100_retrieval_test_sentence.csv

This CSV file contains the caption annotations for the action retrieval testing set and contains 2 columns:

Column NameTypeExampleDescription
narration_idstringP01_11_0Unique ID for the caption (corresponding to the original action).
narrationstringtake plateTranscribed description of the English narration provided by the participant.

Back to Important Files

EPIC_100_train_missing_timestamps_narrations.csv

This CSV file contains the narration IDs of all EPIC-KITCHENS-55 videos in the training set which do not have a narration timestamp (see here for more details). This file has one column:

Column NameTypeExampleDescription
narration_idstringP01_09_660Unique ID for the segment as a string with participant ID and video ID.

Back to Important Files

EPIC_100_validation_missing_timestamps_narrations.csv

This CSV file contains the narration IDs of all EPIC-KITCHENS-55 videos in the validation set which do not have a narration timestamp (see here for more details). This file has one column:

Column NameTypeExampleDescription
narration_idstringP02_12_293Unique ID for the segment as a string with participant ID and video ID.

Back to Important Files

EPIC_100_unseen_participant_ids.csv

This CSV file contains the list of participant IDs who are unseen during training for use in evaluating the unseen participant metrics.

We have two files for both the validation and test set:

Column NameTypeExampleDescription
participant_idstringP33ID of the participant (unique per participant).

Back to Important Files

EPIC_100_tail_verbs.csv

This CSV file contains the list of verb classes which are considered part of the tail classes. These are the set of smallest classes (i.e. those with fewest instances) that account for 20% of the total number of instances in the training set.

Column NameTypeExampleDescription
verbint10Numeric ID representing the verb class.

Back to Important Files

EPIC_100_tail_nouns.csv

This CSV file contains the list of noun classes which are considered part of the tail classes. These are the set of smallest classes (i.e. those with fewest instances) that account for 20% of the total number of instances in the training set.

Column NameTypeExampleDescription
nounint56Numeric ID representing the noun class.

EPIC_100_video_info.csv

This CSV file contains information about each video in the dataset.

Column NameTypeExampleDescription
video_idstrP01_01ID of the video.
durationfloat201.134Duration of the video in seconds.
fpsfloat50.0FPS of the video.
resolutionstr1920x1080Resolution of the video width x height.

Back to Important Files

back to top

Additional Information

Frame indexing

All frame indices are 0 indexed.

File Downloads

Due to the size of the dataset we provide a script for downloading parts of the dataset which can be found here. If you wish to download the extension only (i.e. you have already downloaded EPIC-KITCHENS-55) the following command can be run:

python epic_downloader.py --extension-only

If you wish to download the whole dataset, the following command can be run:

python epic_downloader.py

See the README for more information.

Automatic Annotations Download

We also provide automatic annotations in the form of object masks extracted through the use of MaskRCNN and hand-object BBoxes from ddshan/Hand_Object_Detector (CVPR 2020).

The masks can be downloaded from data.bris and a supporting library is available from this repo.

The hand-object bboxes can be downloaded from data.bris and a supporting library is available from this repo.

Differences to EPIC-Kitchen-55

Updated Annotations

Whilst videos from EPIC-KITCHENS-55 are used within EPIC-KITCHENS-100 some of the annotations have been modified to improve the quality of the annotations. Additionally, with EPIC-KITCHENS-100, the verb/noun classes have been updated to cover the annotations from the new videos. Because of this, the annotations from EPIC-KITCHENS-55 cannot be used for EPIC-KITCHENS-100.

Missing Narration Timestamps

Due to the differences in the annotation pipeline between EPIC-KITCHENS-100 and EPIC-KITCHENS-55, it was impossible to assign the narration timestamp to every action. Because of this, there are actions within EPIC_100_train.csv and EPIC_100_validation.csv which do not have timestamp narrations and are thus marked with NaN within the dataframes.

Pickle Files

We also provide pickle files for all of the main train/val/test csvs for ease of use. These files require python 3.5+ and pandas 1.0.0+ to read. The pickle files are automatically tagged with the commit hash and version for version control purposes which can be found in python using the following commands:

>>> import pandas as pd
>>> train = pd.read_pickle('EPIC_100_train.pkl')
>>> train._metadata
{'commit_hash': 'ce7a0fb', 'version_number': '1.0.0'

showing that this version of the EPIC_100_train.pkl came from commit hash ce7a0fb and version number 1.0.0.

back to top

License

All files in this dataset are copyright by us and published under the Creative Commons Attribution-NonCommerial 4.0 International License, found here. 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.

back to top

Disclaimer

EPIC-KITCHENS-55 and EPIC-KITCHENS-100 were collected as a tool for research in computer vision, however, it is worth noting that the dataset may have unintended biases (including those of a societal, gender or racial nature).

back to top

Changelog

Please see the release history for the changelog.

Current Version 1.1.0.

back to top