Research map / Detection and segmentation / Open-set and few-shot detection
Analysing Domain Shift Factors between Videos and Images for Object Detection
Vicky S. Kalogeiton, Vittorio Ferrari, Cordelia Schmid
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 |
|---|---|---|---|
| replace automatic PRE boxes with ground-truth boxes for video training (ablation) | DPM · Test VOC | mAP gap: 15.7% -> 11.8% | Improved |
| replace automatic PRE boxes with ground-truth boxes for video training (ablation) | R-CNN · Test VOC | mAP gap: 27.3% -> 18.5% | Improved |
| add unique-sample equalization (remove near-duplicate frames) (ablation) | DPM · Test VOC | mAP gap: 11.8% -> 8.1% | Improved |
| add unique-sample equalization (ablation) | DPM and R-CNN · Test YTO | mAP gap: 8.8% for DPM and 5.7% for R-CNN | Hurt |
| add Gaussian blur to VOC training samples (ablation) | DPM and R-CNN · Test VOC | mAP gap: 7.0% DPM, 8.1% R-CNN | Improved |
| add motion blur to VOC training samples (ablation) | DPM and R-CNN · Test VOC | mAP gap: 5.1% DPM, 6.3% R-CNN | Improved |
| add motion blur to VOC training samples (ablation) | R-CNN · Test YTO | mAP gap: widens by 2.7% to 8.4% | Hurt |
| add Gaussian blur to VOC training samples (ablation) | DPM and R-CNN · Test YTO | mAP gap | No clear effect |
| equalize aspect distributions of the two training sets (ablation) | R-CNN · Test VOC | mAP gap: closing by 4.3% to 2.0% | Improved |
| equalize aspect distributions of the two training sets (ablation) | R-CNN · Test YTO | mAP gap: reduced to 1.5%, bridged by 6.9% | Improved |
| equalize all factors on ILSVRC 2015 (ablation) | R-CNN · ILSVRC IMG/VID | mAP gap: below 1.5% | Improved |
| add unique-sample equalization on ILSVRC (4x more samples than VOC) (ablation) | R-CNN · ILSVRC test images | mAP gap | No clear effect |
Closest papers by meaning
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.87)
- Task-Specific Inconsistency Alignment for Domain Adaptive Object Detection (CVPR 2022 · similarity 0.87)
- Learning Transformations To Reduce the Geometric Shift in Object Detection (CVPR 2023 · similarity 0.86)
- Adapting Object Detectors via Selective Cross-Domain Alignment (CVPR 2019 · similarity 0.86)
- Benchmarking Object Detectors under Real-World Distribution Shifts in Satellite Imagery (CVPR 2025 · similarity 0.86)
- ContextShift: A Controlled Benchmark for Context Dependence in Object Detection (NeurIPS 2026 · 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.
