Author(s)

Yangjia Zhou

Corresponding Author

Yangjia Zhou

Abstract

The visual information contained in the product is complex, and many kinds of information overlap with each other, so a single category label can't fully describe it. With the great development of AI (Artificial Intelligence) theory and application, the accuracy and stability of image recognition algorithm based on DL (Deep Learning) have been greatly improved, and it has gradually met the needs of various visual application scenarios. In this paper, we use computer vision and DL technology to study how to automatically identify products in images. In chapter 1, we propose an efficient and concise target detection network: DPF PN-net (dual path fusion feature pyramid constructive network). A DPFM(Dual Path Fusion Module) is used to circularly perform two-way feature fusion on the three-layer feature map to increase the reuse of fused features, and a product qualification detection method based on improved Mask R-CNN is proposed. Experimental results show that the improved Mask R-CNN algorithm in this paper has high recognition rate, and the detection speed is better than that of the original Mask R-CNN algorithm.

Keywords

Computer vision, DL, Automatic product detection