Real-Time Detection of Underwater Debris Using YOLOv11: An Optimized Approach

The debris produced by humans is generally dumped into the ocean or rivers, seriously threatening the marine environment and species. Detecting debris under oceans and rivers is essential to mitigating its dangerous effects on ecosystems and preserving environmental balance. The study’s primary goal is to create an efficient, fast and lightweight debris detection model that works effectively in challenging environments such as low-light conditions and complex underwater terrain. We propose a fine-tuned YOLOv11n model for underwater debris detection under challenging environments, such as complex underwater terrain and low light conditions. Underwater garbage datasets consisting of 11 classes are utilized for training the model, and the dataset is increased with data augmentation techniques to train the model well on complex underwater images. The mAP of our fine-tuned YOLOv11 model is 90%, with an F1-Score of 89% and a Recall of 93%. The number of parameters has been reduced by 30%, with the model size being only 5.5 MB. Future work will focus on expanding the dataset to increase its diversity and integrating the model with real-time applications for underwater debris detection.