Agent Skills
› aiming-lab/AutoResearchClaw
› cv-detection
cv-detection
GitHub提供目标检测任务的架构选择、训练配方及基准测试最佳实践，适用于YOLO、DETR等模型开发。
Trigger Scenarios
进行物体检测任务
选择检测网络架构如YOLO或DETR
优化检测模型训练流程
Install
npx skills add aiming-lab/AutoResearchClaw --skill cv-detection -g -y
SKILL.md
Frontmatter
{
"name": "cv-detection",
"metadata": {
"author": "researchclaw",
"version": "1.0",
"category": "domain",
"priority": "5",
"references": "Ren et al., Faster R-CNN, NeurIPS 2015; Carion et al., End-to-End Object Detection with Transformers, ECCV 2020",
"trigger-keywords": "detection,object,bbox,yolo,coco,anchor,faster rcnn",
"applicable-stages": "9,10"
},
"description": "Best practices for object detection tasks. Use when working on COCO, VOC, or detection architectures like YOLO and DETR."
}
Object Detection Best Practice
Architecture families:
- One-stage: YOLO (v5/v8), SSD, RetinaNet, FCOS
- Two-stage: Faster R-CNN, Cascade R-CNN
- Transformer: DETR, DINO, RT-DETR
Training recipe:
- Use pre-trained backbone (ImageNet)
- Multi-scale training and testing
- IoU threshold: 0.5 for mAP50, 0.5:0.95 for mAP
- Use FPN for multi-scale feature extraction
- Focal loss for class imbalance in one-stage detectors
Standard benchmarks:
- COCO val2017: ~37 mAP (Faster R-CNN R50), ~51 mAP (DINO Swin-L)
- Pascal VOC: ~80 mAP50 (Faster R-CNN)
Version History
- e2e23c9 Current 2026-07-25 07:48
