LiliMeng / revnet-video

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

########################################################### Moments in Time
###########################################################

The Moments in Time Dataset http://moments.csail.mit.edu/

Mathew Monfort, Bolei Zhou, Sarah Adel Bargal, Alex Andonian, Tom Yan, Kandan Ramakrishnan, Lisa Brown, Quanfu Fan, Dan Gutfruend, Carl Vondrick, Aude Oliva

########################################################### Relevant files: ########################################################### README.txt (this file) license.txt (the license file, this must be included) trainingSet.csv (the training set annotations) validationSet.csv (the validation set annotations) moments_categories.txt (class index order) training/ (folder containing training set mp4 videos) validation/ (folder containing validation set mp4 videos)

########################################################### training/ ########################################################### The directory contains mp4 videos used for training. For example, the video filename "blocking/getty-karate-video-id635808620_4.mp4" is located at "training/blocking/getty-karate-video-id635808620_4.mp4"

########################################################### validation/ ########################################################### The directory contains mp4 videos used for validation.

########################################################### moments_categories.txt ########################################################### Contains each class label and index used for training

########################################################### trainingSet.csv and validationSet.csv ########################################################### Comma-seperated csv containing the video list and annotated label for the traning and validation sets. The files contains the following fields:

  • filename: Unique filename for each video.
  • label: Annotated label for each video
  • annotation positive responses: Number of workers that agree with annotated label
  • annotation negative responses: Number of workers that disagree with annotated label

###########################################################

All cached videos can also be manually downloaded from http://data.csail.mit.edu/soundnet/actions3/. For example, http://data.csail.mit.edu/soundnet/actions3/dancing/yt-5of6ffo6-Bw_205.mp4 for the training video "dancing/yt-5of6ffo6-Bw_205.mp4".

About

License:MIT License


Languages

Language:Python 97.1%Language:Shell 2.9%