Research map / Detection and segmentation / CNN features
End-to-End Multi-Object Detection with a Regularized Mixture Model
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 |
|---|---|---|---|
| add MCM loss to mixture NLL (removing duplicates) (ablation) | D-RMM Sparse R-CNN · COCO val | AP: 35.6 -> 47.0 | Helped |
| replace Sparse R-CNN regression+classification losses with mixture NLL+MCM (D-RMM) | Sparse R-CNN R50 · COCO val | AP: +2.0 (45.0 -> 47.0) | Helped |
| apply NMS to the NLL-only mixture model (ablation) | mixture NLL without MCM · COCO val | AP: 35.6 -> 42.7 | Helped |
| winner-take-all merging of confidences on NLL-only model (ablation) | mixture NLL without MCM · COCO val | AP: 35.6 -> 45.8 (still below MCM 47.0) | Helped |
| MCM with bipartite matching instead of simple max matching (ablation) | D-RMM · COCO val | AP: 46.9 vs 47.0 | No clear effect |
| add objectness-based mixture coefficient vs class probability only in confidence (ablation) | D-RMM · COCO val | ECE / reliability: objectness score histogram closer to confidence than class probability alone | Helped |
| add 11 Group-DETR query groups to Sparse R-CNN (ablation) | Sparse R-CNN · COCO val | AP: 45.0 -> 45.9 (less than ours 47.0) | Helped |
| copy-paste GT duplicates with NMS on Sparse R-CNN (ablation) | Sparse R-CNN · COCO val | AP: 45.2 -> 46.1 at 2 copies, 45.3 at 4 copies (with NMS); drops to 17.4 without NMS | Mixed |
| apply D-RMM to AdaMixer | AdaMixer · COCO test-dev | AP: +1.4 (R101), +1.8 (R50) | Helped |
| apply D-RMM to Sparse R-CNN with Swin-T backbone | Sparse R-CNN · COCO val | AP: +2.5 (47.4 -> 49.9) | Helped |
Closest papers by meaning
- Dense Distinct Query for End-to-End Object Detection (CVPR 2023 · similarity 0.87)
- End-to-End Object Detection With Fully Convolutional Network (CVPR 2021 · similarity 0.86)
- What Makes for End-to-End Object Detection? (ICML 2021 · similarity 0.86)
- MS-DETR: Efficient DETR Training with Mixed Supervision (CVPR 2024 · similarity 0.84)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · similarity 0.84)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.83)
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.