Research map / Detection and segmentation / Open-set and few-shot detection
Continual Adaptation: Environment-Conditional Parameter Generation for Object Detection in Dynamic Scenarios
Deng Li, Aming Wu, Yang Li, Yaowei Wang, Yahong Han
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 domain-aware adapter and alignment for test-time adaptation (ablation) | Direct-Test Faster R-CNN · Diverse-Weather | Avg mAP: 27.9 -> 29.5 | Improved |
| add conditional diffusion parameter generation on top of adapter and alignment (ablation) | full method · Diverse-Weather | Avg mAP: 29.5 -> 30.7 | Improved |
| add conditional parameter generation (ablation) | full method · COCO->COCO-C ResNet-50 | Avg mAP: 22.4 -> 23.0 | Improved |
| use dual-path domain-aware adapter instead of simple LoRA (ablation) | full method · Diverse-Weather | Avg mAP: 28.9 -> 30.7 | Improved |
| use class-centered OT alignment instead of KL divergence alignment (ablation) | full method · Diverse-Weather | Avg mAP: 30.1 -> 30.7 | Improved |
| use class-centered OT alignment instead of KL divergence alignment (ablation) | full method · COCO->COCO-C ResNet-50 | Avg mAP: 22.3 -> 23.0 | Improved |
| continual adaptation with proposed method on Night Clear domain | Diverse-Weather · Night Clear | mAP: 35.5 -> 35.3 | No clear effect |
| continual adaptation on COCO-C source-domain forgetting | ResNet-50 · COCO val (source) | mAP: 44.2 -> 42.4 | Mixed |
| parameter-generated adapter vs SKIP | COCO->COCO-C · COCO->COCO-C Swin-T | Avg mAP: 21.6 -> 22.5 | Improved |
Closest papers by meaning
- Test-Time Adaptive Object Detection with Foundation Model (NeurIPS 2025 · similarity 0.88)
- What How and When Should Object Detectors Update in Continually Changing Test Domains? (CVPR 2024 · similarity 0.88)
- Domain Adaptive Object Detection via Dynamic Causal Refinement (ICML 2026 · similarity 0.88)
- Generalized Diffusion Detector: Mining Robust Features from Diffusion Models for Domain-Generalized Detection (CVPR 2025 · similarity 0.86)
- C-LoRA: Continual Low-Rank Adaptation for Pre-trained Visual Models (NeurIPS 2026 · similarity 0.86)
- PALM: Pushing Adaptive Learning Rate Mechanisms for Continual Test-Time Adaptation (AAAI 2025 · 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.
