Published Paper Details:

Underwater imaging faces significant challenges due to poor visibility, low contrast, color distortion, and light scattering, which hinder the accurate detection of objects such as obstacles and marine life. This project proposes a deep learning-based system that integrates advanced image enhancement techniques with object detection models to improve underwater scene understanding. The enhanced images are fed into a convolution neural network (CNN) or transformer-based detection architecture, such as YOLO or Faster R-CNN, trained to identify and localize both underwater obstacles and fish species. The image enhancement stage leverages methods such as histogram equalization, white balancing, and deep learning-based restoration to recover visual clarity. Experimental results on standard underwater datasets demonstrate improved detection accuracy and robustness under varying visibility conditions, making this approach highly beneficial for autonomous underwater vehicles (AUVs), marine biology research, and underwater navigation systems.