Research map / Vision-language and multimodal / Cross-modal learning
When Pedestrian Detection Meets Multi-Modal Learning: Generalist Model and Benchmark Dataset
Yi Zhang, Wang Zeng, Sheng Jin, Chen Qian, Ping Luo, Wentao Liu
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 the MAF token for adaptive modality weighting (ablation) | MMPedestron encoder · LLVIP (fusion) | AP: 61.7 -> 68.2 | Improved |
| add the MAA token on top of MAF (ablation) | MMPedestron encoder · LLVIP (fusion) | AP: 68.2 -> 68.9 | Improved |
| add multi-modal training stage after RGB pretraining (ablation) | MMPedestron · LLVIP (fusion) | AP: 50.0 -> 72.6 | Improved |
| add multi-modal training stage after RGB pretraining (ablation) | MMPedestron · InOutDoor (fusion) | AP: 36.8 -> 65.7 | Improved |
| finetune on COCO-Persons instead of direct evaluation | MMPedestron · COCO-Persons val | AP: 68.2 -> 71.1 | Improved |
| use only 10% of PEDRo training data for finetuning (ablation) | MMPedestron · PEDRo test (Event) | AP: 73.3 -> 72.7 | No clear effect |
| swap Co-DINO head for Faster R-CNN head (ablation) | MMPedestron · LLVIP (fusion) | AP: 72.6 -> 66.8 | Hurt |
| unified shared encoder instead of per-modality branches | MMPedestron · CrowdHuman | AP: 95.4 -> 97.1 | Improved |
Closest papers by meaning
- Robust Pedestrian Detection with Uncertain Modality (AAAI 2026 · similarity 0.88)
- Weakly Aligned Cross-Modal Learning for Multispectral Pedestrian Detection (ICCV 2019 · similarity 0.88)
- Generalizable Pedestrian Detection: The Elephant in the Room (CVPR 2021 · similarity 0.88)
- Towards Versatile Pedestrian Detector with Multisensory-Matching and Multispectral Recalling Memory (AAAI 2022 · similarity 0.88)
- Causal Mode Multiplexer: A Novel Framework for Unbiased Multispectral Pedestrian Detection (CVPR 2024 · similarity 0.87)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.84)
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.
