GaryFer/smart-parking-weights

Smart Parking UPeU — Model Weights

Trained model weights for parking slot detection comparing five architectures on the Smart Parking UPeU v4 dataset (Juliaca, Peru).

Models included

• YOLOv8s
• YOLOv11s
• YOLOv12s
• RT-DETR-L
• Faster R-CNN (ResNet-50 FPN)

Dataset

3 classes: libre, ocupado, no_disponible
3,072 training images / 293 validation / 146 test

Results (mAP@0.5, mean ± SD, 10 runs)

Model	mAP@0.5	FPS
YOLOv8s	0.9948 ± 0.0002	205.5 ± 6.4
YOLOv11s	0.9947 ± 0.0001	161.2 ± 5.7
YOLOv12s	0.9946 ± ?	94.2 ± 26.9
RT-DETR-L	0.9946 ± 0.0002	41.1 ± 0.6
Faster R-CNN	0.9925 ± 0.0003	26.9 ± 0.9

Citation

@article{yunganina2026smartparking,
  title={Smart Parking Detection...},
  author={Yunganina Mamani, Gary Fernando},
  year={2026}
}