CLASSES = {0:"body", 1:"face", 2:"frame", 3:"text"}     # confirmed by user

def get_yolo_predictions(yolo_model,img_path:str):
    results = yolo_model.predict(source=img_path, device='cuda')
    dets = {"body":[], "face":[], "frame":[], "text":[]}
    for box in results[0].boxes:
        c  = int(box.cls[0])
        x1, y1, x2, y2 = map(int, box.xyxy[0])
        dets[CLASSES[c]].append([x1,y1,x2,y2])
    return dets