CUHK team won the ImageNet challenge on object detection in videos
ImageNet Large Scale Visual Recognition Challenge (ILSVRC) is one of the most important grand challenges in computer vision. There are more than 70 top computer vision groups participated in ILSVRC 2015. The CUHK team (CUvideo) including Prof. Ouyang Wanli, Prof. Li Hongsheng, Prof. Wang Xiaogang and five PhD students from EE department participated in ILSVRC 2015 and won the challenge of object detection from videos.
CUvideo achieved the mean Averaged Precision (mAP) of 67.8%, while the team in the second place only obtained the mAP of 35.9%. The challenge required detecting 30 object categories from one thousand videos and CUvideo won 28 categories out of them. The CUHK team also ranked NO.2 in object detection in images (Microsoft was the winner). It required detection 200 object categories from 40,000 images. This was the second time for CUHK to participated in this ILSVRC. In ILSVRC2014, the CUHK team ranked the second in object detection in images, while Google was the Winner.