基于YOLOv5s网络的垃圾分类和检测

目的 为了实现垃圾自动按类处理,通过研究基于视觉的垃圾检测与分类模型,实现对垃圾的自动识别和检测.方法 采用YOLOv5s网络作为垃圾检测与分类的模型,在自制垃圾分类数据集上对网络进行训练,利用训练好的YOLOv5s网络提取不同种类垃圾图像的特征和位置信息,实现垃圾的分类与检测.结果 在真实场景中进行了测试,基于YOLOv5s的垃圾分类检测模型可以有效识别6种不同形态的垃圾,检测mAP值为99.38％,测试精度为95.34％,目标检测速度达到6.67FPS.结论 实验结果表明,基于YOLOv5s网络的垃圾分类检测模型在不同光照、视角等条件下,检测准确率高,鲁棒性好、计算速度快.同时,有助于促进垃圾处理公司实现智能分拣,提高工作效率.

Garbage Classification and Detection Based on YOLOv5s Network

To automatically dispose garbage according to its category, the automatic recognition of garbage can be realized by doing research on the vision based garbage detection and classification model. YOLOv5s network is taken as a garbage classification and detection model and trained on the self-made garbage data set. The trained YOLOv5s network extracts the features and location information from different kinds of garbage images, and then recognizes and detects different garbage in the image. The performance of the trained YOLOv5s is validated in a real situation. Garbage classification and detection model which was based on YOLOv5s can identify six different kinds of garbage effectively. The mAP of the trained YOLOv5s is 99.38%, the recognition accuracy is 95.34%, and the speed of target detection reaches 6.67FPS.The experiment results show that the garbage classification model based on YOLOv5s has high accuracy, good robustness and fast speed under different situations such as illumination, camera angle, etc. At the same time, this study can help enterprises to realize intelligent sorting and improve the efficiency.