 The proposed method is a generic framework for enhancing underwater images prior to running SLAM algorithms. It uses a generative adversarial network, GAN, to improve the quality of the images, reducing the need for manual post-processing. Additionally, it utilizes knowledge distillation to compress the GAN model, making it more efficient and suitable for real-time applications. This allows for faster and more reliable SLAM results. This article was authored by Zhichang Zhang, Zhichao XIN, Zhibi Niu, and others.