상세정보
학술대회
Deep learning-based Feature Compression for Video Coding for Machine
- 발행일
- 202201
- 출처
- International Workshop on Advanced Image Technology (IWAIT) 2022 (SPIE 12177), pp.1-5
- 협약과제
- 21HH4800, [전문연구실] 기계를 위한 영상 부호화, 서정일
- 초록
- We previously trained the compression network via optimization of bit-rate and distortion (feature domain MSE) [1]. In this paper, we propose feature map compression method for Video Coding for Machine (VCM) based on deep learning-based compression network that joint training for optimizing both compressed bit rate and machine vision task performance. We use bmshij2018-hyperporior model in the CompressAI [2] as the compression network and compress the feature map which is the output of stem layer in the Faster R-CNN X101-FPN network of Detectron2 [3]. We evaluated the proposed method by Evaluation Framework for MPEG VCM. The proposed method shows the better results than VVC of MPEG VCM Anchor.
- KSP 제안 키워드
- Bit Rate, Compression method, Evaluation Framework, Faster r-cnn, Feature Compression, Feature Map, Learning-based, Video coding, deep learning(DL), joint training, machine vision