AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

CMD: Self-Supervised 3D Action Representation Learning with Cross-Modal Mutual Distillation

Yunyao Mao, Wengang Zhou, Zhenbo Lu, Jiajun Deng, Houqiang Li

ECCV 2022 · Vision-language and multimodal › Cross-modal learning · approach: Unimodal-to-multimodal representation transfer · 70 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 cross-modal mutual distillation (joint and motion) (ablation)MoCo v2 baseline, BiGRU · NTU-60 x-sub lineartop-1 accuracy: 76.1 -> 79.8Improved
add distillation among all three modalities (ablation)MoCo v2 baseline · NTU-60 x-sub KNNtop-1 accuracy: 63.4 -> 70.6Improved
make distillation bidirectional instead of joint<-bone only (ablation)CMD · NTU-60 x-sub lineartop-1 accuracy: 76.5 -> 77.7Improved
make distillation bidirectional instead of joint<-motion only (ablation)CMD · NTU-60 x-sub lineartop-1 accuracy: 78.9 -> 79.8Improved
make distillation bidirectional instead of joint<-motion only (ablation)CMD · NTU-60 x-sub KNNtop-1 accuracy: 64.8 -> 68.7Improved
make distillation bidirectional among three modalities (ablation)CMD · NTU-60 x-sub KNNtop-1 accuracy: 66.5 -> 70.6Improved
make distillation bidirectional among three modalities (ablation)CMD · NTU-60 x-sub lineartop-1 accuracy: 78.8 -> 79.4Improved
increase number of neighbors K beyond 8192 (ablation)CMD · NTU-60 x-sub lineartop-1 accuracyNo clear effect
replace CrosSCLR-B positive mining with CMDBiGRU encoder · NTU-60 x-sub KNNtop-1 accuracy: 66.1 -> 70.6Improved
replace CrosSCLR-B positive mining with CMDBiGRU encoder · PKU-II transfer from NTU-120top-1 accuracy: 52.8 -> 57.0Improved
replace CrosSCLR-B positive mining with CMDBiGRU encoder · NTU-60 x-sub semi-supervised 1%top-1 accuracy: 48.6 -> 50.6Improved
add multi-modal ensembling at test timeCMD · NTU-60 x-sub lineartop-1 accuracy: 79.8 -> 84.1Improved

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