M2-V Video De-Duplication

PI: Zhu Li

The goal of this project is to develop novel deep learning algorithms for video segment hashing and identification to support efficient and accurate duplicates identification and removal from phones and cloud storages.  The project will utilize methods supporting transcoding and editing effects, scalable hash from Fisher Vector aggregation of the conv features, contrastive learning with MLP on Fisher Vectors, and deduplication search acceleration.  Deliverables are algorithm research and evidences in publications, MPEG/JVET standards contributions, and software PoC.