Open-Closed Depth Hand Dataset (OCDH)

closed-hand samples:
open-hand samples:

The Open-Closed Depth Hand Dataset (OCDH) is composed of samples of a pair of classes representing open or closed hand poses. The project was carried out by researchers and students from the Institute of Computing of Federal University of Alagoas, Brazil. Each sample is given as a depth (scalar) image and a binarized image. A large and diversified hand pose image dataset composed by 160,000 samples was collected from 20 individuals (15 men and 5 women). From each subject, depth images of the whole body were collected using a Kinect v1 sensor at 640 × 480 resolution. To improve diversity, individuals were asked to vary its global body pose as much as possible, and capture was performed in environments with different backgrounds. From the depth images, a cropped hand was extracted and stored in the dataset. As the distance from the hand to the sensor is variable, the cropped image resolution varies greatly. The database was originally collected to evaluate the methodology presented in the paper "Bruno Lima, Givanildo L. N. Júnior, Lucas Amaral, Thales Vieira, Bruno Ferreira and Tiago Vieira, 'Real-time hand pose tracking and classification for natural human-robot control', In International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISAPP 2019."

A preprint version of the paper is available here.
The database can be downloaded here.

Our database is free to use for research purposes. Please cite our database in your work:

author={Bruno Lima and Givanildo {N. Júnior} and Lucas Amaral and Thales Vieira and Bruno Ferreira and Tiago Vieira},
title={Real-time Hand Pose Tracking and Classification for Natural Human-Robot Control},
booktitle={Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 5: VISAPP,},

For questions, please contact Prof. Thales Vieira.