Research map / 3D vision / Human pose estimation
Human Pose Estimation in Extremely Low-Light Conditions
Sohyun Lee, Jaesung Rim, Boseung Jeong, Geonu Kim, Byungju Woo, Haechan Lee, Sunghyun Cho, Suha Kwak
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 LSBN and neural-style LUPI with paired well-lit images | CPN, ResNet-50 · ExLPose LL-A | AP@0.5:0.95: 25.4 -> 32.7 | Improved |
| Add LSBN to pose-only Baseline-all (ablation) | CPN, ResNet-50 · ExLPose LL-A | AP@0.5:0.95: 25.4 -> 30.1 | Improved |
| Add LUPI to LSBN (ablation) | CPN, ResNet-50 · ExLPose LL-A | AP@0.5:0.95: 30.1 -> 32.7 | Improved |
| Use LUPI without LSBN (ablation) | CPN, ResNet-50 · ExLPose LL-A | AP@0.5:0.95: 25.4 -> 24.0 | Hurt |
| Use aligned low-light and well-lit pairs for training (ablation) | LSBN + LUPI · ExLPose LL-A | AP@0.5:0.95: 30.2 -> 32.7 | Improved |
| Use accurate low-light pose labels instead of unlabeled low-light data (ablation) | LSBN + LUPI · ExLPose LL-A | AP@0.5:0.95: 4.6 -> 32.7 | Improved |
| Match teacher neural styles rather than feature maps directly (ablation) | LSBN + LUPI · ExLPose LL-A | AP@0.5:0.95: 30.4 -> 32.7 | Improved |
| Apply LUPI from C1 through R4 rather than only C1 (ablation) | LSBN + LUPI · ExLPose LL-A | AP@0.5:0.95: 29.7 -> 32.7 | Improved |
| Send LUPI gradients only from teacher to student (ablation) | LSBN + LUPI · ExLPose LL-A | AP@0.5:0.95: 31.6 -> 32.7; 29.5 -> 32.7 | Improved |
| Scale low-light inputs to average channel intensity 0.4 (ablation) | Ours · ExLPose LL-A | AP@0.5:0.95: 14.5 -> 32.7 | Improved |
| Use image enhancement before pose estimation | Baseline-all · ExLPose LL-E | AP@0.5:0.95: 14.3 -> 8.3 (LLFlow); 14.3 -> 12.5 (LIME) | Hurt |
| Use image enhancement before pose estimation | Baseline-all · ExLPose LL-N | AP@0.5:0.95: 33.8 -> 35.2 (LLFlow); 33.8 -> 38.3 (LIME) | Improved |
Closest papers by meaning
- Domain-Adaptive 2D Human Pose Estimation via Dual Teachers in Extremely Low-Light Conditions (ECCV 2024 · similarity 0.92)
- UDAPose: Unsupervised Domain Adaptation for Low-Light Human Pose Estimation (CVPR 2026 · similarity 0.89)
- Simultaneously-Collected Multimodal Lying Pose Dataset: Enabling In-Bed Human Pose Monitoring (TPAMI 2022 · similarity 0.85)
- SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation (AAAI 2024 · similarity 0.84)
- PoseIRM: Enhance 3D Human Pose Estimation on Unseen Camera Settings via Invariant Risk Minimization (CVPR 2024 · similarity 0.84)
- Towards Robust Event-guided Low-Light Image Enhancement: A Large-Scale Real-World Event-Image Dataset and Novel Approach (CVPR 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.
