Thank you for downloading the DGAZE dataset. The DGAZE dataset is a driver eye gaze tracking database which has object-level and point-level annotations. After downloading the zip file, please check that the md5 checksum of DGAZE_dataset.zip is equal to c781af1e832c281d02782a3f0c248667 This dataset is released in a compressed format and has the following structure: The main folder contains sub folders for drivers 2,3,5,7,8,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24 and a road_view folder Each driver folder contains a csv file with the frame drop rate and a folder with 112 video clips corresponding to the 112 video clips in the road_view folder. The road_view folder also contains the framewise groundtruth gaze location for each clip. The first 9 video clips correspond to the 9 callibration points. They have associated groundtruth npy files which have 100 2D points. The rest of the videos are road videos. They have associated txt files in the following format: frame# x1 y1 x2 y2 which give the endpoints of the ground truth object bounding box. The center of the bounding box is the ground truth gaze point. To use the database, please clone this github repo: https://github.com/duaisha/DGAZE and follow the instructions in the README to extract the data for training purpose. This code will adjust the road view videos to have the same framerate as NEED TO WRITE THE CORRECT STRUCTURE FOR THIS Visit our project page here: http://cvit.iiit.ac.in/research/projects/cvit-projects/dgaze If you are using this dataset, please cite this paper: @inproceedings{isha2020iros, title={DGAZE: Driver Gaze Mapping on Road}, author={Dua, Isha and John, Thrupthi Ann and Gupta, Riya and Jawahar, CV}, booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2020}, year={2020} } For any enquiries, please contact thrupthi.ann@research.iiit.ac.in