Construction material visual inventory method based on lightweight deep neural network

• Main Authors: LIU JIAN, DENG XI, CAO XINYU, LIU SHOUSONG, WANG KAI, YANG YINGMING, QIAN QI
• Format: Patent
• Language: Chinese; English
• Published: 08.12.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention relates to the technical field of computer vision, in particular to a construction material visual inventory method based on a lightweight deep neural network, and the method comprises the steps: collecting a building material data set, marking a steel bar data set through manual combination with semi-automatic marking, and dividing the steel bar data set into a training set and a test set; a lightweight neural network model is designed on the basis of Improved ShufflNet v2; training the lightweight neural network model by using the constructed training set, and preliminarily testing the performance of the lightweight neural network model by using the test set; performing channel pruning on the trained lightweight neural network model based on a BN layer, simplifying the lightweight neural network model, and testing the performance of the pruned lightweight neural network model by using the test set again; and detecting the reinforcing steel bar image by using the trained and pruned lightweight