Underwater Object Detection Based on Improved EfficientDet

Intelligent detection of marine organism plays an important part in the economy, and it is significant to detect organisms quickly accurately a complex environment for intelligence equipment. The existing object models do not work well underwater. This paper improves structure EfficientDet detector proposes EfficientDet-Revised (EDR), which new model. Specifically, MBConvBlock reconstructed by adding Channel Shuffle module enable exchange information between channels feature layer. fully connected layer attention removed convolution used cut down amount network parameters. Enhanced Feature Extraction constructed multi-scale fusion enhance extraction ability different objects. results experiments demonstrate that mean average precision (mAP) proposed method reaches 91.67% 92.81% on URPC dataset Kaggle dataset, respectively, better than other models. At same time, processing speed 37.5 frame per second (FPS) can meet real-time requirements. It provide useful reference underwater robots perform tasks such as intelligent grasping.