Research map / Detection and segmentation / CNN features
MutDet: Mutually Optimizing Pre-training for Remote Sensing Object Detection
Ziyue Huang, Yongchao Feng, Qingjie Liu, Yunhong Wang
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 mutual enhancement module feeding enhanced features to decoder (ablation) | MutDet · DIOR-R 12 epochs | AP50: 62.8 -> 65.5 | Improved |
| add auxiliary siamese head for calibration instead of distillation (ablation) | MutDet · DIOR-R (pretrain DOTA-v1.0) | AP50: 69.4 -> 70.7 | Improved |
| calibrate with encoder feature distillation (ablation) | MutDet · DIOR-R (pretrain DOTA-v1.0) | AP50: 69.4 -> 69.6 | No clear effect |
| calibrate with decoder cross distillation (ablation) | MutDet · DIOR-R (pretrain DOTA-v1.0) | AP50: 69.4 -> 69.9 | Improved |
| add contrastive alignment loss to DETReg baseline (ablation) | DETReg · DIOR-R 36 epochs | AP50: 67.9 -> 68.5 | Improved |
| add encoder contrastive alignment loss without siamese head (ablation) | MutDet · DIOR-R 36 epochs | AP50: 69.5 -> 69.4 | No clear effect |
| use MutDet pre-training instead of DETReg | ARS-DETR R50 · DIOR-R 10% labeled data | AP50: 50.8 -> 56.9 | Improved |
| pre-train on DOTA then fine-tune on DOTA with prior methods | UP-DETR, AlignDet, DETReg · DOTA-v1.0 test | AP50: 72.9 -> 72.6 | Hurt |
| use stronger Swin-Tiny backbone and larger RSDet4 pre-training data | MutDet vs DETReg · DIOR-R 36 epochs | AP50: 73.2 -> 73.7 | No clear effect |
| use MutDet pre-training instead of DETReg on other DETR detectors | Rotated D-DETR · DIOR-R | AP50: 41.4 -> 45.8 | Improved |
| AlignDet pre-training instead of none | ARS-DETR R50 · DIOR-R 12 epochs | AP50: 55.4 -> 54.3 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DETReg — DETReg: Unsupervised Pretraining With Region Priors for Object Detection (CVPR 2022) · better on 9 of 9 shared results
- UP-DETR — UP-DETR: Unsupervised Pre-Training for Object Detection With Transformers (CVPR 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SM3Det: A Unified Model for Multi-Modal Remote Sensing Object Detection (AAAI 2026 · similarity 0.90)
- AlignDet: Aligning Pre-training and Fine-tuning in Object Detection (ICCV 2023 · similarity 0.88)
- Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community (AAAI 2025 · similarity 0.86)
- SARDet-100K: Towards Open-Source Benchmark and ToolKit for Large-Scale SAR Object Detection (NeurIPS 2024 · similarity 0.85)
- Benchmarking Object Detectors under Real-World Distribution Shifts in Satellite Imagery (CVPR 2025 · similarity 0.85)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.85)
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.
