AI Paper Atlas

Research map / Video / Video segmentation and tracking

SMC++: Masked Learning of Unsupervised Video Semantic Compression

Yuan Tian, Xiaoyue Ling, Cong Geng, Qiang Hu, Guo Lu, Guangtao Zhai

TPAMI 2025 · Video › Video segmentation and tracking · approach: Self-supervised video representation learning · 19 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add masked learning instead of MoCoV3 contrastive objective (ablation)SMC · UCF101-TSM @0.02bppTop1 accuracy: 75.20% -> 76.48%Improved
add masked learning instead of MoCoV3 contrastive objective (ablation)SMC · MOT17-ByteTrack @0.01bppMOTA: 67.62% -> 70.84%Improved
add non-semantic suppression (NSS) loss (ablation)SMC · UCF101-TSMTop1 accuracy: 69.57% -> 72.96%Improved
add semantic feature as GMM condition (ablation)SMC · UCF101-TSMTop1 accuracy: 71.38 -> 72.96Improved
use masking ratio 90% instead of 50% (ablation)SMC · UCF101-TSM @0.02bppTop1 accuracy: 73.85 -> 76.48Improved
use masking ratio 90% instead of 95% (ablation)SMC · UCF101-TSM @0.02bppTop1 accuracy: 76.08 -> 76.48No clear effect
predict DINOv2 features instead of raw pixels (ablation)SMC · UCF101-TSMTop1 accuracy: 76.48 -> 77.00 at 0.02bpp; 89.11 -> 89.51 at 0.06bppMixed
add semantic compensation operation (ablation)SMC · UCF101-TSM @0.02bppTop1 accuracy: 16% drop when removedImproved
add LPIPS loss term (ablation)SMC · UCF101-TSM @0.02bppTop1 accuracy: about 15% drop when removedImproved
add GAN loss term (ablation)SMC · UCF101-TSM @0.02bppTop1 accuracy: about 5% drop when removedImproved
add masked motion objective with RGB difference target (ablation)SMC-adv · VOS DAVIS2017 @0.02bppJ&F: 79.30 -> 79.76Improved
add Blue-Tr module (ablation)SMC-adv+Motion · VOS DAVIS2017 @0.02bppJ&F: 79.76 -> 81.06Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy