Training a custom object detection model - YOLO v10 vs RT-DETR vs custom?