Research map / Detection and segmentation / CNN features
RT-DETRv4: Painlessly Furthering Real-Time Object Detection with Vision Foundation Models
Zijun Liao, Yian Zhao, Xin Shan, Yu Yan, Chang Liu, lei lu, Xiangyang Ji, Jie Chen
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 DSI and GAM (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 53.8 -> 54.3 | Improved |
| add DSI alone (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 53.8 -> 53.9 | No clear effect |
| add GAM on top of DSI (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 53.9 -> 54.3 | Improved |
| add DSI+GAM (ablation) | RT-DETRv2-L · COCO val2017 36 epochs | AP: 52.1 -> 52.6 | Improved |
| add DSI+GAM (ablation) | D-FINE-L · COCO val2017 36 epochs | AP: 53.1 -> 53.4 | Improved |
| add DSI on RT-DETRv2-L (ablation) | RT-DETRv2-L · COCO val2017 36 epochs | AP75: 56.7 -> 56.4 | Hurt |
| align only AIFI output F5 instead of backbone features (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 53.7 -> 54.3 | Improved |
| add backbone alignment to F5 alignment (hybrid) (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 54.3 -> 53.8 | Hurt |
| use linear projector instead of MLP (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 54.2 -> 54.3 | No clear effect |
| use DINOv3 teacher instead of MAE (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 53.9 -> 54.3 | Improved |
| use DINOv3 teacher instead of DINOv2 (ablation) | DEIM-L · COCO val2017 36 epochs | AP: 54.1 -> 54.3 | Improved |
| use cosine similarity loss instead of MSE (ablation) | DEIM-M · COCO val2017 90 epochs | AP: 52.7 -> 53.7 | Improved |
Closest papers by meaning
- Frozen-DETR: Enhancing DETR with Image Understanding from Frozen Foundation Models (NeurIPS 2024 · similarity 0.88)
- Hybrid Proposal Refiner: Revisiting DETR Series from the Faster R-CNN Perspective (CVPR 2024 · similarity 0.88)
- DETRs Beat YOLOs on Real-time Object Detection (CVPR 2024 · similarity 0.87)
- MI-DETR: An Object Detection Model with Multi-time Inquiries Mechanism (CVPR 2025 · similarity 0.87)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- KD-DETR: Knowledge Distillation for Detection Transformer with Consistent Distillation Points Sampling (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.
