An object detection network for locating and recognizing cluttered garbage
Gaoqiang Ji, Shuxin Liu · 2023
The research on garbage classification have made remarkable achievements in the latest work. However, due to the complexity of the location elements of the garbage background, there are still many difficulties in the accuracy of garbage location detection. For complex backgrounds, it is easy to cause false detection and missed detection. In response to the above problems, we explored the improved YOLOv5n network , added the ContextAggregation module and replaced GIOU with Focal-EIOU, and replaced C3 with C3SAC to improve feature extraction for contextual information. The improved algorithm has been proved by experiments to increase mAP by 2.3%. The accuracy rate of Plastic bowl, Power banks, Cardboard boxes, Tea leaves, Plastic hangers and so on has reached more than 80%.