Research map / Detection and segmentation / Open-set and few-shot detection
Context Aware Grounded Teacher for Source Free Object Detection
Tajamul Ashraf, Rajes Manna, Partha Sarathi Purkayastha, Taha Tariq, Janibul Bashir
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 Semantic Augmentation (SA) on top of RCM and AT base (ablation) | Grounded Teacher · Cityscapes→Foggy ablation (VGG16) | mAP: 30.4 -> 41.8 | Improved |
| add Semantic-Aware Loss (SAL) on top of RCM and AT base (ablation) | Grounded Teacher · Cityscapes→Foggy ablation (VGG16) | mAP: 30.4 -> 44.2 | Improved |
| add SAL to RCM+SA to form full GT (ablation) | Grounded Teacher · Cityscapes→Foggy ablation (VGG16) | mAP: 41.8 -> 50.8 | Improved |
| add SA to RCM+SAL to form full GT (ablation) | Grounded Teacher · Cityscapes→Foggy ablation (VGG16) | mAP: 44.2 -> 50.8 | Improved |
| pair related classes (SA) instead of random Mixup pairing (ablation) | Grounded Teacher · Cityscapes→Foggy ablation (VGG16) | mAP: 46.4 -> 50.8 | Improved |
| lower augmentation probability from 1.0 to 0.5 (ablation) | Semantic Augmentation · Cityscapes→Foggy ablation (VGG16) | mAP: 47.5 -> 50.8 | Improved |
| use SAL with inter-class relations instead of diagonal-only class-level weights (ablation) | Grounded Teacher · Cityscapes→Foggy ablation (VGG16) | mAP: 47.1 -> 50.8 | Improved |
| add regularization term to SAL (ablation) | SAL · Cityscapes→Foggy ablation (VGG16) | mAP: 47.8 -> 50.8 | Improved |
| add expert branch guidance (ablation) | Grounded Teacher · DDSM to INBreast and DDSM to RSNA | R@0.3: +0.10 | Improved |
| apply GT source-free adaptation vs source-only model | Faster R-CNN VGG-16 · Cityscapes→Foggy Cityscapes (SFOD, VGG16) | mAP: 18.4 -> 50.8 | Improved |
| apply GT adaptation vs source-only on surgical shift | Detector · InSeg1 to InSeg2 | R@1.0 and AUC: 0.25 -> 0.30 R@1.0; 0.03 -> 0.21 AUC | Improved |
| apply GT adaptation vs source-trained model | Detector · Cityscapes to BDD100k (source-free) | mAP: 20.2 -> 29.1 | Improved |
Closest papers by meaning
- Simplifying Source-Free Domain Adaptation for Object Detection: Effective Self-Training Strategies and Performance Insights (ECCV 2024 · similarity 0.90)
- Cross-Domain Adaptive Teacher for Object Detection (CVPR 2022 · similarity 0.89)
- Dynamic Retraining-Updating Mean Teacher for Source-Free Object Detection (ECCV 2024 · similarity 0.89)
- Periodically Exchange Teacher-Student for Source-Free Object Detection (ICCV 2023 · similarity 0.89)
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.88)
- CAT: Exploiting Inter-Class Dynamics for Domain Adaptive Object Detection (CVPR 2024 · 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.
