UWV-Yolox: A Deep Learning Model for Underwater Video Object Detection

Underwater video object detection is a challenging task due to the poor quality of underwater videos, including blurriness and low contrast. In recent years, Yolo series models have been widely applied detection. However, these perform poorly for blurry low-contrast videos. Additionally, they fail account contextual relationships between frame-level results. To address challenges, we propose model named UWV-Yolox. First, Contrast Limited Adaptive Histogram Equalization method used augment Then, new CSP_CA module proposed by adding Coordinate Attention backbone representations objects interest. Next, loss function proposed, regression jitter loss. Finally, optimization optimize results utilizing relationship neighboring frames in improving performance. evaluate performance our model, We construct experiments on UVODD dataset built paper, select [email protected] as evaluation metric. The UWV-Yolox reaches 89.0%, which 3.2% better than original Yolox model. Furthermore, compared with other models, has more stable predictions objects, improvements can be flexibly models.