Convolutive Neural Networks for Automatic Categorization of bovine carcasses through image
artificial intelligence, classification, typification
The automatic classification of the conformation and finishing of bovine carcasses are tools for obtaining information to the production chain, not only in the usual scope of weight and yield, but also in the qualitative aspects, helping in the management of the enterprises, enabling improvement in the production rates established with based on data generated by expert systems, as well as allowing a fair valuation of the marketed product. The objective of this study is to automatically typify beef carcasses in a refrigerator using different architectures of Convolutional Neural Networks (CNN), in order to increase the objectivity of the process. The formation of the database took place from July 4 to 29, 2022 in commercial slaughterhouses/refrigerators that are partners to the project, located in the state of Pará - PA, where detailed carcass information such as hot carcass weight, sex and age were collected to perform the classification and typification of the carcasses, as well as the 2D imaging of about 2600 bovine carcasses for further typification and processing, corresponding to Step 1: Experimental Collection: from an identification and separation of carcasses according to their sex and age classification and hot carcass weight. During weighing, the half carcasses were imaged in the side view. It is intended to carry out new collections to increase the sample n. The punctual evaluation of the Typification took place after the evaluation. Step 2: it will occur after the removal of the frames, and thus, determine how the evaluators were trained and after the completion of the configuration via image. Step 3: Database Pre-Processing and Filming Processing, from a database created by the organization, using the R-CNN Mask. And finally, Step 4: Automatic Classification, in this future phase, with the neural network trained, new entries will be created for the validation of the network.