Tree species identification method based on improved YOLOv7

The paper presents a natural tree species recognition methods based on YOLOv7. We propose a new small target detection layer based on the YOLOv7 network, use the improved Mosaic-8, and introduce the attention mechanism. On the basis of not affecting the detection speed of YOLOv7, we improve the detection accuracy. Experiments show that the method has stronger learning ability, and accuracy than other algorithms under the same conditions.