Published Paper Details:

Underwater visual environments suffer from severe image degradation due to light absorption, scattering, color distortion, and turbidity, which significantly affects the performance of automated fish species recognition systems. To address these challenges, this paper presents an integrated deep learning framework for underwater fish species detection by combining diffusion-based image enhancement with a lightweight classification model. Initially, a diffusion probabilistic restoration model is employed to enhance underwater images by improving color fidelity, suppressing noise, and restoring contrast. The enhanced images are then classified using a MobileNetV2-based convolutional neural network optimized for real-time and resource-constrained deployment. Experimental evaluation is conducted using standard image quality metrics such as UIQM, UCIQE, PSNR, and SSIM, along with classification metrics including accuracy, precision, recall, and F1-score. The results demonstrate significant improvements in both visual quality and species recognition accuracy compared to conventional CNN-based approaches trained on raw underwater images. The proposed framework offers a robust and scalable solution for applications in aquaculture monitoring, marine biodiversity analysis, and underwater robotic vision systems.