Try new yolos

[robmarkcole]

• https://blog.roboflow.com/how-to-train-yolox-on-a-custom-dataset/ -> got an issue with the data loading, possibly due to pascal-VOV format error? Gave up
• https://blog.roboflow.com/train-yolor-on-a-custom-dataset/
• https://blog.roboflow.com/how-to-train-scaled-yolov4/
• https://blog.roboflow.com/how-to-train-a-custom-mobile-object-detection-model/ (can be converted to tflite)

[robmarkcole]

YoloR gives great result without even augmentation:

 P           R      mAP@.5      mAP@.5:.95
0.681     0.833       0.831          0.408

[robmarkcole]

Scaled yolo-v4

P               R     mAP@.5      mAP@.5:.95:
0.443       0.775       0.705          0.281

[robmarkcole]

Yolov4-tiny

precision = 0.69, recall = 0.59, F1-score = 0.64 

mAP@0.50 = 0.663356

[robmarkcole]

done