The Kinetics Dataset could be downloaded from the following link:
After all the videos were downloaded, resize the video to the short edge size of 256, then prepare the csv files for training, validation, and testing set as train.csv
, val.csv
, test.csv
. The format of the csv file is:
path_to_video_1 label_1
path_to_video_2 label_2
path_to_video_3 label_3
...
path_to_video_N label_N
-
Please download the dataset and annotations from dataset provider.
-
Download the frame list from the following links: (train, val).
-
Extract the frames at 30 FPS. (We used ffmpeg-4.1.3 with command
ffmpeg -i "${video}" -r 30 -q:v 1 "${out_name}"
in experiments.) Please put the frames in a structure consistent with the frame lists.
Please put all annotation json files and the frame lists in the same folder, and set DATA.PATH_TO_DATA_DIR
to the path. Set DATA.PATH_PREFIX
to be the path to the folder containing extracted frames.