YOLOv8n Model Release v1.0.0
We are excited to announce the release of our YOLOv8n-Traffic model trained on a filtered COCO dataset.
Highlights of this release
- The model was trained for 100 epochs, ensuring a robust learning from the dataset.
- We used a filtered version of the COCO dataset, focusing on specific classes to improve detection accuracy in those areas.
Performance Metrics
- Mean Average Precision (mAP@.50): 63.8%
- Mean Average Precision (mAP50-95): 46.31%
- Precision: 71.25%
- Recall: 56.59%
Known Issues
- Only trained in 100 epochs
- Imbalance image object count
Future Plans
- Fine Tune the model
- Better performance on stitched image