Sample dataset

This release contains a demonstration of what the actual test set will look like.
It contains input videos, ground truth images, and companion files for 10 documents.

Any feedback welcome!

Sample dataset without json files and frames

This is a pre-release, not the final sample dataset

This sample dataset contains video sequences and associated ground-truth images illustrating the kind of challenge participants will face. Extra information like the bounding box of the object to reconstruct on a given frame of the video and the shape of the target image, will be provided soon in a later release.

(Target image shape can be deduced from the ground-truth file image shape.)