A 3D-CNN based video hashing method
Jiande Sun, Haifeng Qi, Jing Li, Wenbo Wan, Qiang Wu · 2018
Most video hashing algorithms have the common pipeline, which consists of feature extraction and hash mapping. The performance of video hash is usually promoted via the improvement in one or both of aspects. In this paper, a learningbased video feature is used, which is obtained via a 3D-CNN model. The 3DCNN-based features can represent both spatial and temporal information of videos, as 3D convolutions used in 3DCNN can capture the motion information through multiple adjacent frames. A video hashing algorithm is proposed based on the 3DCNN-based feature, which is defined as CNNF. In addition, the hash length optimization method is used to get the approximately optimal hash length in hash mapping stage of the proposed algorithm. Since the feature extraction and hash length optimization are independent to hash mapping algorithms, several classical hashing algorithms are adopted to verify the improvements of these two aspects via the video copy detection task. Experiments demonstrate the performance of the proposed CNNFHash algorithm.