Lipstick quality inspection-oriented transfer learning target detection method

The invention discloses a lipstick quality inspection-oriented transfer learning target detection method, and belongs to the technical field of deep learning target detection. A transfer learning target detection method for lipstick quality inspection optimizes a network structure on the basis of a YOLO v5 network structure, and more specifically comprises a construction method of a lipstick data set, a target detection model YOLO v5X network structure and an introduction transfer learning method. According to the method, the generalization performance of lipstick defect detection can be effectively improved, and the requirement of the target detection model for the data volume is effectively reduced. 本发明公开了一种面向口红质检的迁移学习目标检测方法，属于深度学习目标检测技术领域。一种面向口红质检的迁移学习目标检测方法，以YOLO v5网络结构为基础，对网络结构进行优化，更具体说，包括有口红数据集的构建方法、目标检测模型YOLO v5X网络结构、引入迁移学习方法；本发明能够有效地提高口红缺陷检测的泛化性能，有效降低了目标检测模型对于数据量的需求。