SSD-based Lightweight Recyclable Garbage Target Detection Algorithm

Abstract

Reducing garbage pollution to the environment will improve the quality of people’s living environment and achieve a green cycle of life. In response to the existing traditional garbage recycling and classification, which requires a lot of wasted human and material resources, this paper proposes a garbage detection network with higher accuracy for garbage identification and localization by improving the SSD network. Firstly, the backbone network is improved to generate multiple scales of feature maps through the Resnet network as the basis for subsequent detection. In addition, the SE module is introduced to optimize the features needed for the model and weaken the environment’s interference in the detection model. At the input image size of 320×320, the map of the improved model in terms of accuracy is 90.18%, which is an improvement of 1.82% compared to the original network. It can also meet the demand for real-time detection in terms of detection rate. Finally, compared with the common model, the improved model has better results in terms of detection accuracy and detection rate.