Kinetics is a collection of large-scale, high-quality datasets of URL links of up to 650,000 video clips that cover 400/600/700 human action classes, depending on the dataset version. The videos include human-object interactions such as playing instruments, as well as human-human interactions such as shaking hands and hugging. Each action class has at least 400/600/700 video clips. Each clip is human annotated with a single action class and lasts around 10 seconds.
The Kinetics project publications can be found here: https://deepmind.com/research/open-source/kinetics.
Note that it may not be safe to train / test across different versions of the dataset, for example the k400 validation set is largely part of the k700 training set so results will not be meaningful.
5th of May: fixed k400/train/part_120.tar.gz, it was a tar file before
10th of December: add two downloader scripts for datasets automatic setup. (k400_downloader.sh and k700_2020_downloader.sh)
13th of September. K400: replaced corrupted mountain climber validation file and made available 1300+ replacement videos for existing corrupted training videos from various classes. K600: added list of links for held out test set.
4th of August 2021 -- replaced corrupted videos in the kinetics-700-2020 test set (these were typically shorter than 9s as well). There are still 5% of the videos in the test set that are shorter than 9s, but genuinely so (e.g. because they are like that in youtube).
CVDF currently hosts the videos in the Kinetics-400 and Kinetics-700-2020 datasets.
git clone https://github.com/cvdfoundation/kinetics-dataset.git
cd kinetics-dataset
This will create two directories, k400 and k400_targz. Tar gzips will be in k400_targz, you can delete k400_targz directory after extraction.
bash ./k400_downloader.sh
bash ./k400_extractor.sh
The train/val/test splits are subdivided into many files. The lists of links to video files can be found here:
https://s3.amazonaws.com/kinetics/400/train/k400_train_path.txt
https://s3.amazonaws.com/kinetics/400/val/k400_val_path.txt
https://s3.amazonaws.com/kinetics/400/test/k400_test_path.txt
It is easy to obtain a specific split (e.g. train), by:
bash download.sh k400_train_path.txt
Then, extract *.tar.gz files by:
bash extract.sh k400_train_path.txt
The links/annotations can be found under the annotation subfolders:
https://s3.amazonaws.com/kinetics/400/annotations/train.csv
https://s3.amazonaws.com/kinetics/400/annotations/val.csv
https://s3.amazonaws.com/kinetics/400/annotations/test.csv
A readme file can be found in:
http://s3.amazonaws.com/kinetics/400/readme.md
News: users found ~1400 corrupted videos. A replacement for the vast majority can be found here:
https://s3.amazonaws.com/kinetics/400/replacement_for_corrupted_k400.tgz
git clone https://github.com/cvdfoundation/kinetics-dataset.git
cd kinetics-dataset
This will create two directories, k600 and k600_targz. Tar gzips will be in k600_targz, you can delete k600_targz directory after extraction.
bash ./k600_downloader.sh
bash ./k600_extractor.sh
The train/val/test splits are subdivided into many files. The lists of links to video files can be found here:
https://s3.amazonaws.com/kinetics/600/train/k600_train_path.txt
https://s3.amazonaws.com/kinetics/600/val/k600_val_path.txt
https://s3.amazonaws.com/kinetics/600/test/k600_test_path.txt
The links/annotations can be found under the annotation subfolders:
https://s3.amazonaws.com/kinetics/600/annotations/train.txt
https://s3.amazonaws.com/kinetics/600/annotations/train.csv (incomplete)
https://s3.amazonaws.com/kinetics/600/annotations/val.txt
https://s3.amazonaws.com/kinetics/600/annotations/val.csv (incomplete)
https://s3.amazonaws.com/kinetics/600/annotations/test.csv
https://s3.amazonaws.com/kinetics/600/annotations/kinetics600_holdout_test.csv
A readme file can be found in:
http://s3.amazonaws.com/kinetics/600/readme.md
git clone https://github.com/cvdfoundation/kinetics-dataset.git
cd kinetics-dataset
This will create two directories, k700-2020 and k700-2020_targz. Tar gzips will be in k700-2020_targz, you can delete k700-2020_targz directory after extraction.
bash ./k700_2020_downloader.sh
bash ./k700_2020_extractor.sh
The train/val/test splits are subdivided into many files. The lists of links to video files can be found here:
https://s3.amazonaws.com/kinetics/700_2020/train/k700_2020_train_path.txt
https://s3.amazonaws.com/kinetics/700_2020/val/k700_2020_val_path.txt
https://s3.amazonaws.com/kinetics/700_2020/test/k700_2020_test_path.txt
The links/annotations can be found under the annotation subfolders:
https://s3.amazonaws.com/kinetics/700_2020/annotations/train.csv
https://s3.amazonaws.com/kinetics/700_2020/annotations/val.csv
https://s3.amazonaws.com/kinetics/700_2020/annotations/test.csv
A readme file can be found in:
http://s3.amazonaws.com/kinetics/700_2020/K700_2020_readme.txt
We also host annotations for AVA-Kinetics and Countix, which both use Kinetics-700 videos.
To download annotations for AVA-Kinetics: https://s3.amazonaws.com/kinetics/700_2020/annotations/ava_kinetics_v1_0.tar.gz
To download annotations for countix: https://s3.amazonaws.com/kinetics/700_2020/annotations/countix.tar.gz
