Research map / Video / Action recognition
Pairwise Contrastive Learning Network for Action Quality Assessment
Mingzhe Li, Hong-Bo Zhang, Qing Lei, Zongwen Fan, Jinghua Liu, Ji-Xiang Du
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 |
|---|---|---|---|
| add PCLN branch (ablation) | basic regression network · AQA-7 | average SRC: 0.8504 -> 0.8721 | Improved |
| add consistency constraint (ablation) | basic regression + PCLN · AQA-7 | average SRC: 0.8721 -> 0.8795 | Improved |
| add PCLN branch (ablation) | basic regression network · MTL-AQA FSP | SRC: 0.8745 -> 0.8788 | Improved |
| add PCLN branch (ablation) | basic regression network · MTL-AQA ESP | SRC: 0.9095 -> 0.9196 | Improved |
| add consistency constraint (ablation) | basic regression + PCLN · MTL-AQA FSP | SRC: 0.8788 -> 0.8798 | Improved |
| add consistency constraint (ablation) | basic regression + PCLN · MTL-AQA ESP | SRC: 0.9196 -> 0.9230 | Improved |
| increase batch size from 8 to 32 (28 to 496 video pairs) (ablation) | PCLN · AQA-7 | average SRC: 0.8276 -> 0.8795 | Improved |
| increase batch size from 32 to 64 (496 to 2016 pairs) (ablation) | PCLN · AQA-7 Diving | SRC: 0.8697 -> 0.8189 | Hurt |
| increase batch size from 8 to 64 (28 to 2016 pairs) (ablation) | PCLN · MTL-AQA FSP | SRC: 0.8729 -> 0.8798 | Improved |
| increase batch size from 8 to 64 (28 to 2016 pairs) (ablation) | PCLN · MTL-AQA ESP | SRC: 0.9094 -> 0.9230 | Improved |
| predict execution score then multiply by difficulty (ESP) instead of final score directly (FSP) | PCLN · MTL-AQA | SRC: 0.8798 -> 0.9230 | Improved |
| use temporal encoder features instead of P3D 3DCNN features (ablation) | basic regression network · AQA-7 diving_007 | predicted score stability: 83.07 fixed over 10 tests, ground truth 83.25 | Improved |
Closest papers by meaning
- Group-Aware Contrastive Regression for Action Quality Assessment (ICCV 2021 · similarity 0.88)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.87)
- Learning Skill-Attributes for Transferable Assessment in Video (NeurIPS 2025 · similarity 0.81)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.80)
- GAIA: Rethinking Action Quality Assessment for AI-Generated Videos (NeurIPS 2024 · similarity 0.80)
- PTM-VQA: Efficient Video Quality Assessment Leveraging Diverse PreTrained Models from the Wild (CVPR 2024 · similarity 0.79)
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.
