Text this: IA-YOLO: A Vatica Segmentation Model Based on an Inverted Attention Block for Drone Cameras