Fins play an important role in zebrafish maneuvering; However, their action mechanism is still poorly understood. Therefore, we use a conjunctive method of maneuvering photography and deep learning to complete the image reconstruction of zebrafish pectoral and caudal fins by improving the YOLO model and the U-Net model. By training on 16,000 images and testing on 700 images, the final mDice values achieved for fin reconstruction were 95.91% for the left pectoral fin, 95.94% for the right pectoral fin, and 94.23% for the caudal fin. Based on this, this paper portrays the deformation characteristics of zebrafish's left and right pectoral and caudal fins during maneuvering. In the range of 9.01° to 50.68° of zebrafish turning angle, the average areas of the left and right pectoral fins were 8.40 mm2 and 8.85 mm2, and the average change of the open angles was 29.97° and 24.38°. When the zebrafish was turning counterclockwise, the change in the open angle of the left pectoral fin was more obvious, and the average area of extension was smaller than that of the right pectoral fin. Conversely, the left fin is larger than the right fin. During rotation, the open angle of the zebrafish caudal fin does not change significantly, but its mean area increases with the rotation angle. In summary, the results of this paper have a positive impact on the mechanism of zebrafish fins in fast movements, which can be studied in depth.