YOLOv3-darknet with Adaptive Clustering Anchor Box for Garbage Detection in Intelligent Sanitation
Wei Hong Cui, Wei Zhang, Juli Green, Xu Zhang, Xiang Yao · 2019 3rd International Conference on Electronic Information Technology and Computer Engineering (EITCE) · 2019
In recent years, as people’s requirements for living environment are getting higher and higher, many countries have written environmental protection into relevant laws. How to find and clean up garbage in a timely manner is particularly important. Manually searching garbage is not only time consuming, but also requires a lot of manpower. Deep learning has achieved good results in the fields of object detection and semantic segmentation with a good ability to identify these uncertain features just like garbage. In this paper, we capture the garbage image on the road in real time by sanitation vehicle. The YOLOv3-darknet model which is based on adaptive clustering anchor box are used for garbage detection. Multi-category confrontation training is performed on objects that are frequently misdetected. Combined with other auxiliary ranging devices, the latitude and longitude information of the garbage is determined, so that the garbage can be cleaned with a target. Experiments show that our model has achieved good results in garbage detection. The detection time of a single image on the GTX1080ti is 60ms.