Chest-X-ray-Abnormalities-Detection

VinBigData Chest X-ray Abnormalities Detection (Dataset)[https://www.mediafire.com/file/xwiio9bezjfc1l2/yolo_data.zip/file]:

• Already convert to YOLO format
• Dataset folder structure similar to YOLOv5

Usage:

Train args:

''' python train.py --img 640 --batch 4 --epochs 5 --data ../dataset.yaml --weights yolov5x.pt --workers 0
'''

''' python train.py --img 640 --batch 8 --epochs 30 --data ../dataset.yaml --weights runs/train/exp9/weights/last.pt --workers 0 '''

Detect args:

''' python detect.py --weight runs/train/exp19/weights/last.pt --img 640 --conf 0.25 --iou 0.4 --source H:\test_set_chest_xray\test --save-txt --save-conf --exist-ok '''

Note:

• workers should set to 0 (I'm using RTX 2070)

RESULT:

• exp15 - 0.064
• exp17 - 0.116
• exp19 - 0.084