ml_project_requirements

metrics: accuracy. recall. precision. f1_score. specificty auc_score

loss
box_loss?
DFL loss?
cls_loss?

#kfold cross validation

must: no of images are equal or not for each class

#if required perform augmentation

later: compare with other variants of yolo including the latest 


now
1. list out no of all classes in training
2. seperate the images into train, val, test - from annotated dataset - train 4168(301000 to 305225), val 1191(from 305226 to 306432), test 596(remaining)
3. redo the script.py for each folder
4. simplify performance metrics

later
1. try out training yolov10, yolov6



references citation format

yolo model documentation reading