Ship detection method based on enhanced Yolov4-tiny algorithm

The invention discloses a ship detection method based on an enhanced Yolov4-tiny algorithm, and belongs to the technical field of ship information detection, and the method comprises the steps: taking a Yolov4-tiny network as a basic model, and constructing an improved Yolov4-tiny network model; the method specifically comprises the following steps: taking Darknet53-tiny as a trunk feature extraction network, and taking an improved three-dimensional BiFPN structure as a feature enhancement part; adding a target detection head network; a Soft-NMS algorithm is adopted to remove redundant detection frames, and a Pytorch framework is used for implementation; inputting the collected SSDD data set into the improved network model, and carrying out iterative training until the classification network converges; and inputting a test set sample into the trained improved network model, outputting a result, and checking a detected ship effect picture. According to the invention, the ship detection capability can be improv