LSMDC Video Description Dataset

LSMDC stands for Large Scale Movie Description Challenge. The dataset contains 118,081 short video clips extracted from 202 movies. Each video is accompanied by subtitles, some of which are extracted from the movie scripts, and some are transcribed through DVS (oral description service for the visually impaired).

The validation set contains 7,408 video clips, and the evaluation is performed on a test set consisting of 1,000 movie videos that do not overlap with the training and validation sets.