Table 2

An overview of human-machine collaborative video compression methods in literature. MBID, MBHD, and SBMD respectively represent multi-bitstream independent decoding, multi-bitstream hierarchical decoding, and single-bitstream multi-head decoding.

CategoryAuthorPresented TaskCore Method
MBID[50]Video RetrievalFeature extrAction + CDVS + CNN
[211]Video RetrievalRate-accuracy optimization + affine motion compensation
[10]Class Identification, Object RecognitionComprising Multiple autoencoders
MBHD[197]Action RecognitionConditional deep generation network
[82]Action RecognitionSemantic information + feature Laddering Framework
[114]Object DetectionConditional semantic compression + interlayer frame prediction
[64]Object DetectionEnd-to-end learnable video codec + conditional coding
[39]Object DetectionConventional + DNN video compression
[85]Action RecognitionLearned semantic representation + end-to-end optimize
[93]Object Detection, Pose Estimation, Action Recognition, Object SegmentationStatic Object characteristic + dynamic motion clue
[170]Action Recognition, Multiple Object Tracking, Object SegmentationTraditional codec + DNN
[171]Action Recognition, Multiple Object Tracking, Object SegmentationSemantic-Mining-then-Compensation + masked image modeling
[4]Object DetectionCuboidal feature descriptor
SBMD[207]Action RecognitionTask-driven optimization
[160]Action Recognition, Object Detection, Object Tracking, Object SegmentationTemporal context + cross-domain motion
[ViewLarge]
This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal