Research map / Detection and segmentation / Open-set and few-shot detection
CLIP the Gap: A Single Domain Generalization Approach for Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| initialize the detector with CLIP pre-trained weights (ablation) | Faster R-CNN, ResNet101 · Day Foggy | mAP: 32.0 -> 37.5 | Helped |
| add CLIP text-based classification loss (ablation) | CLIP-initialized detector · Dusk Rainy and Night Rainy versus other weather conditions | mAP: 31.0 -> 31.3; 15.7 -> 16.3; 37.5 -> 36.9; 37.0 -> 36.0 | Mixed |
| add CLIP attention pooling (ablation) | CLIP-initialized detector with text loss · Day Foggy | mAP: 36.9 -> 37.7 | Helped |
| add CLIP attention pooling (ablation) | CLIP-initialized detector with text loss · Night Clear | mAP: 36.0 -> 35.9 | Hurt |
| add relevant weather and time-of-day semantic augmentations (ablation) | CLIP detector with text loss and attention pooling · Night Rainy | mAP: 16.7 -> 18.7 | Helped |
| replace no augmentation with random feature perturbations (ablation) | CLIP detector with text loss and attention pooling · Night Rainy | mAP: 16.7 -> 15.3 | Hurt |
| use unrelated CLIP prompts instead of weather-relevant prompts (ablation) | CLIP detector with text loss and attention pooling · Dusk Rainy | mAP: 32.3 -> 30.2 | Hurt |
| use CLIP-based semantic augmentation instead of no augmentation (ablation) | CLIP detector with text loss and attention pooling · Day Foggy | mAP: 37.7 -> 38.5 | Helped |
| use the proposed method instead of S-DGOD | object detector · Dusk Rainy | mAP: 28.2 -> 32.3 | Helped |
| use the proposed method instead of S-DGOD | object detector · Day Clear | mAP: 56.1 -> 51.3 | Hurt |
| use the proposed method instead of S-DGOD | object detector · Night Clear | mAP: 36.6 -> 36.9 | Helped |
Closest papers by meaning
- Towards Single-Source Domain Generalized Object Detection via Causal Visual Prompts (NeurIPS 2025 · similarity 0.92)
- Object-Aware Domain Generalization for Object Detection (AAAI 2024 · similarity 0.91)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.90)
- Style Evolving along Chain-of-Thought for Unknown-Domain Object Detection (CVPR 2025 · similarity 0.89)
- Unbiased Faster R-CNN for Single-source Domain Generalized Object Detection (CVPR 2024 · similarity 0.89)
- Domain Generalization in CLIP via Learning with Diverse Text Prompts (CVPR 2025 · similarity 0.88)
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.