YOLO Experiments

This repo contains my YOLO training and validation experiments for Person detection.

What’s inside

• Colab notebooks used for training and testing
• Different model sizes (YOLO11n, YOLO11s)
• Different image sizes (832, 960)
• Albumentations / Roboflow dataset setup
• Saved runs and results

Dataset

• 1 class: Person
• Train/valid split from Roboflow
• Augmentations applied during training

Latest validation results

YOLO11s @ 832

• Precision: 0.878
• Recall: 0.879
• mAP50: 0.915
• mAP50–95: 0.405

YOLO11n @ 960

• Precision: 0.865
• Recall: 0.837
• mAP50: 0.893
• mAP50–95: 0.399

Conclusion: YOLO11s (832) is slightly better overall.