Research map / Detection and segmentation / Open-set and few-shot detection
Task-conditioned Domain Adaptation for Pedestrian Detection in Thermal Imagery
My Kieu, Andrew D. Bagdanov, Marco Bertini, Alberto del Bimbo
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 task-conditioning layers on detection heads (TC Det) (ablation) | YOLOv3 thermal detector · KAIST night | log-average miss rate: 16.69 -> 10.31 | Improved |
| add task-conditioning layers on detection heads (TC Det) (ablation) | YOLOv3 thermal detector · KAIST day and night | log-average miss rate: 31.06 -> 27.11 | Improved |
| condition residual groups instead of no conditioning (ablation) | YOLOv3 thermal detector · KAIST day and night | log-average miss rate: 31.06 -> 28.69 | Improved |
| condition all residual blocks (ablation) | YOLOv3 thermal detector · KAIST day and night | log-average miss rate: 31.06 -> 31.54 | Mixed |
| add residual group conditioning on top of detection head conditioning (ablation) | TC Det · KAIST day and night | log-average miss rate: 27.11 -> 30.48 | Hurt |
| fine-tune on visible images before thermal (no conditioning) (ablation) | YOLOv3 thermal detector · KAIST day and night | log-average miss rate: 31.06 -> 30.50 | Improved |
| add conditioning during visible stage (ablation) | visible-then-thermal schedule · KAIST day and night | log-average miss rate: 30.50 -> 28.48 | Improved |
| add conditioning during thermal stage (ablation) | visible-then-thermal schedule · KAIST night | log-average miss rate: 15.73 -> 11.03 | Improved |
| add conditioning in both visible and thermal stages (ablation) | visible-then-thermal schedule · KAIST day and night | log-average miss rate: 30.50 -> 27.11 | Improved |
| add conditioning to thermal-only adaptation without visible stage (ablation) | YOLOv3 thermal detector · KAIST night | log-average miss rate: 16.69 -> 12.61 | Improved |
| add conditioning to visible detector (ablation) | YOLOv3 visible detector · KAIST visible test | log-average miss rate: 36.67 -> 34.73 | Improved |
| use TC Det (thermal only) versus Bottom-up adaptation | YOLOv3 · KAIST night | log-average miss rate: 20.50 -> 10.31 | Improved |
Closest papers by meaning
- D3T: Distinctive Dual-Domain Teacher Zigzagging Across RGB-Thermal Gap for Domain-Adaptive Object Detection (CVPR 2024 · similarity 0.87)
- A Robust Learning Approach to Domain Adaptive Object Detection (ICCV 2019 · similarity 0.86)
- Domain-Specific Suppression for Adaptive Object Detection (CVPR 2021 · similarity 0.86)
- Automatic Adaptation of Object Detectors to New Domains Using Self-Training (CVPR 2019 · similarity 0.86)
- Multiple Adverse Weather Conditions Adaptation for Object Detection via Causal Intervention (TPAMI 2022 · similarity 0.86)
- Causal Mode Multiplexer: A Novel Framework for Unbiased Multispectral Pedestrian Detection (CVPR 2024 · similarity 0.86)
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.
