US 11,676,257 B2
Method and device for detecting defect of meal box, server, and storage medium
Yawei Wen, Beijing (CN); Jiabing Leng, Beijing (CN); Minghao Liu, Beijing (CN); Yulin Xu, Beijing (CN); Jiangliang Guo, Beijing (CN); and Xu Li, Beijing (CN)
Assigned to BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed by BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed on Nov. 27, 2019, as Appl. No. 16/697,944.
Claims priority of application No. 201811458919.X (CN), filed on Nov. 30, 2018.
Prior Publication US 2020/0175673 A1, Jun. 4, 2020
Int. Cl. G06T 7/00 (2017.01); B65B 25/00 (2006.01); G06N 3/08 (2023.01)
CPC G06T 7/0004 (2013.01) [B65B 25/001 (2013.01); G06N 3/08 (2013.01); G06T 2207/10004 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A method for detecting defect of a meal box, performed by a server, comprising:
receiving a detection request comprising an image of the meal box sent by a user terminal, wherein the image of the meal box is obtained by an image acquirer of the user terminal; and
performing defect recognition based on the image of the meal box and a defect detection model in response to the detection request;
wherein after performing defect recognition based on the image of the meal box and the defect detection model in response to the detection request, the method further comprises:
obtaining correction information of a defect type and defect location coordinates, and generating new defect detection samples according to the correction information, in which the new defect detection samples are samples with defects corrected by the correction information of the defect type and the defect location coordinates; and
updating the defect detection model based on the new defect detection samples;
wherein the defect detection model is trained by:
training a neural network model based on sample meal box images and defect locations and defect types in the sample meal box images by using a mask region-based convolutional neural network (R-CNN) algorithm, and optimizing network model parameters by combining a loss of the mask R-CNN algorithm with a loss of a faster R-CNN;
wherein the server is determined from a plurality of servers deployed with defect detection models of different detection functions by identifying a meal box shape and a production environment.