Detection dataset designed for Integrated Detection Method of Pedestrian and Face.
The FEEDS (Face pEdestrain dEtection DataSet) dataset is integrated with both face and pedatrian detection bounding boxes, contains 82735 annotated pedestrian bounding boxes and 20771 annotated face bounding boxes.
Our dataset is collected from three different dataset:
- PRW (Person Re-identification in the Wild) dataset.
- WIDER Pedestrian dataset.
- CUHK Person Search (Large Scale Benchmark for Person Search) dataset.
Dataset is formed with images and annotation, annotations can be found in 'annotations' folder, images and the whole dataset can be downloaded from BaiduYun Link (Extract code:e9i9).
Data analyse can be found in Wiki Page.
[1] Zheng L, Zhang H, Sun S, et al. Person re-identification in the wild[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 1367-1376.
[2] Yang S, Luo P, Loy C C, et al. Wider face: A face detection benchmark[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 5525-5533.
[3] Xiao T, Li S, Wang B, et al. End-to-end deep learning for person search[J]. arXiv preprint arXiv:1604.01850, 2016, 1(2).