Research map / Video / Action recognition
Weakly Supervised Micro- and Macro-Expression Spotting Based on Multi-Level Consistency
Wang-Wang Yu, Kai-Fu Yang, Hongmei Yan, Yongjie Li
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use CSCM instead of direct concatenation of image and optical-flow features (ablation) | MC-WES · CAS(ME)^2 | F1-score: 0.207 -> 0.324 | Improved |
| use CSCM instead of CCM (ablation) | MC-WES · CAS(ME)^2 | F1-score: 0.236 -> 0.324 | Improved |
| use CSCM instead of direct concatenation (ablation) | MC-WES · SAMM-LV | F1-score: 0.168 -> 0.212 | Improved |
| use different top-K values per class ([7,9,5]) instead of identical values (ablation) | MC-WES · CAS(ME)^2 | F1-score: 0.182 -> 0.324 | Improved |
| set MaE value lower than ME value in h ([7,9,5] vs [7,7,5]) (ablation) | MC-WES · CAS(ME)^2 | F1-score: 0.238 -> 0.324 | Improved |
| snippet duration 8 instead of 16 frames (ablation) | MC-WES · CAS(ME)^2 | F1-score: 0.153 -> 0.324 | Improved |
| snippet duration 8 instead of 4 frames (ablation) | MC-WES · CAS(ME)^2 | F1-score: 0.211 -> 0.324 | Improved |
| snippet duration 32 instead of 16 or 48 frames (ablation) | MC-WES · SAMM-LV | F1-score: 0.159 -> 0.212 | Improved |
| add L_sc (saliency consistency) alone vs L_sl alone (ablation) | MC-WES · CAS(ME)^2 loss ablation | F1-score: 0.049 -> 0.067 | Improved |
| add L_dc1 and L_dc2 to L_sc, L_sl, L_gl (ablation) | MC-WES · CAS(ME)^2 loss ablation | F1-score: 0.071 -> 0.191 | Improved |
| add L_fc on top of L_sc, L_dc1, L_dc2, L_sl, L_gl (ablation) | MC-WES · CAS(ME)^2 loss ablation | F1-score: 0.191 -> 0.212 | Improved |
| add L_dc3 (video-level distribution consistency) when L_dc2 already used (ablation) | MC-WES · CAS(ME)^2 loss ablation | F1-score: 0.212 -> 0.324 | Mixed |
Closest papers by meaning
- Region-Aware Instance Consistency Learning for Micro-Expression Recognition (CVPR 2026 · similarity 0.87)
- Cross-Task Relation-Aware Consistency for Weakly Supervised Temporal Action Detection (TPAMI 2025 · similarity 0.86)
- Gaussian-Based Instance-Adaptive Intensity Modeling for Point-Supervised Facial Expression Spotting (ICLR 2025 · similarity 0.86)
- Weakly Supervised Video Salient Object Detection (CVPR 2021 · similarity 0.86)
- Weakly Supervised Instance Segmentation for Videos With Temporal Mask Consistency (CVPR 2021 · similarity 0.85)
- Local-Global Multi-Modal Distillation for Weakly-Supervised Temporal Video Grounding (AAAI 2024 · similarity 0.84)
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.
