Ocean engineering, computer graphics, and surveillance all increasingly prioritize underwater photography as a research topic. Because deeper water cannot be reached by longer wavelengths of sunlight, underwater photographs seem blue-green and murky. A big dataset of underwater photographs called enhancement of underwater visual perception (EUVP) is utilized. The cycle generative adversarial networks (CycleGANs) approach is used in this model to improve image quality. This determines an image’s quality depending on the image's global color, content, local texture, and style information. Then, the model compares enhanced image and original image by using standard metrics named structural similarity (SSIM). Conventional algorithms for underwater item identification, person position calculation, and saliency prediction function better when employing more accurate photographs. In order to make the information in photographs easier for viewers to understand or to offer better input for other automated image processing techniques, the model is trained to improve the underwater image.