YOLOv7-Based Land and Underwater Target Detection and Recognition

There is a growing demand for marine resource development. Bionic amphibious robots can replace humans to conduct land and underwater exploration, which has important research significance. Deep learning has developed rapidly in recent years, and many kinds of target detection algorithms have emerged. We select one two-stage target detection algorithm Faster R-CNN, three single-stage target detection algorithms SSD, Centernet, and YOLOv7. We use each of these four algorithms to train the VOC2007 dataset in a deep learning environment. After the training is completed, these four models are evaluated and predicted separately. We find an algorithm that is most suitable for the amphibious robot applicationYOLOv7. Finally, we use the YOLOv7 model to detect the underwater dataset, and the results prove that the model is promising for detecting small underwater targets.