Object Detection Methods
Detect and localize objects in images.
Two-Stage Detectors
R-CNN: selective search + CNN features. Fast R-CNN: share CNN computation. Faster R-CNN: RPN for proposals.
Region proposals + classification.
One-Stage Detectors
YOLO: divide image into grid, predict boxes. SSD: multi-scale feature maps.
Faster than two-stage, slightly less accurate.
Metrics
mAP: mean average precision. IoU: intersection over union. Higher IoU = stricter.
Key Takeaways
- Two-stage: R-CNN family
- One-stage: YOLO, SSD
- mAP evaluates detection performance