Detection-Guided Deep Unfolding for Joint Underwater Image Enhancement and Object Detection

Project Code :TCMAPY2403

Objective

This project aims to develop an integrated underwater object detection system by combining image enhancement with YOLO11m and advanced fusion techniques. It focuses on training and evaluating the model on the URPC2020 dataset using metrics like precision, recall, and mAP over 30 epochs to ensure high detection accuracy. Additionally, the framework explores improvements in robustness and real-time performance for efficient underwater detection.

Abstract

This research aims to enhance underwater images and detect objects within them using a dual approach of image enhancement and object detection. The project leverages a novel fusion technique combining YOLO11m with advanced dynamic fusion methods. These techniques improve the performance of object detection by refining image quality in underwater environments. The YOLO11m model, integrated with Spatial-wise Dynamic Fusion and Cross-Scale Efficient Attention, is trained and evaluated on the URPC2020 dataset for 30 epochs. The proposed system is designed to address the challenges posed by underwater imaging, where lighting and color distortions often obscure visual information. By improving the visibility of underwater objects and refining detection accuracy, this project aims to advance the capabilities of autonomous systems in underwater environments.

Keywords: YOLO11m, Spatial-wise Dynamic Fusion, Cross-Scale Efficient Attention, Image Enhancement, Object Detection, Underwater Imaging, Deep Learning, Computer Vision, URPC2020, Autonomous Systems

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.