SmartTrato: computer vision and artificial intelligence plataform to improve feed management based on animal behaviour

• Grant number: 18/14609-0
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: May 01, 2019
• End date: March 31, 2020
• Field of knowledge: Agronomical Sciences - Animal Husbandry - Animal Nutrition and Feeding
• Principal Investigator: Alan Caio Rodrigues Marques
• Grantee: Alan Caio Rodrigues Marques
• Company: Tech Inovações Tecnológicas para a Agropecuária S/A
• CNAE: Criação de bovinos ; Desenvolvimento de programas de computador sob encomenda ; Atividades profissionais, científicas e técnicas não especificadas anteriormente
• City: Manaus
• Associated scholarship(s): 19/10284-2 - SmartTrato: computer vision and artificial intelligence platform to improve feed management based on animal behaviour, BP.TT; 19/11675-5 - SmartTrato: computer vision and artificial intelligence platform to improve feed management based on animal behavior, BP.TT; 19/11676-1 - SmartTrato: computer vision and artificial intelligence platform to improve feed management based on animal behaviour, BP.TT; 19/09674-0 - SmartTrato: computer vision and artificial intelligence plataform to improve feed management based on animal behaviour, BP.PIPE

Abstract

More than 2/3 of the production cost of intensive cattle systems is represented by diet. Therefore, the economic success of these operations depends on the food distribution for the animals. Currently the food supply adjustment depends on the evaluation of trained people, assigning subjective grades of score reading in the trough, once or twice a day. Climate events are also not used to adjust the diet offer. Using predictive models, minimizing the human error, and improving the food management can help the country to: reduce diet losses, that may reach R$ 500 million/year and improve the feed efficiency (food convertion to meat). This project, named SmartTrato, aims to: 1) to develop algorithms to eliminate the subjectivity of food management, 2) to understand how the machines can be automated and 3) to minimize the human error, reducing diet losses and improve food efficiency, increasing the farm profitability. SmartTrato will characterized the feedlot, climate data, diet chemical composition, trough losses (orts), and animal behavior. The hypothesis of this study is that adjusted feed efficiency for orts and profitability are significantly better using the SmartTrato Method (MS) compared to the Traditional Method (MT). In this PIPE Phase 1 the algorithms will be developed and trained; only in Phase 2 the methods will be compared. Two monitoring systems will be developed with cameras (named TratoCam): i) attached to the feed mixer (MSV) moving next to the trough lines and ii) static cameras (MSE) positioned on the trough line. Both will monitor the diet in the trough, the orts and the animal attitude. Sampling of the ingredients, diet and losses, as well as weighing data for calibration of the algorithms, will be conducted during the project. A local weather station will be used to monitor the climatic events, associating with exogenous climatic data to adjust the predictive algorithms. The animals will be monitored simultaneously by SmartTrato and individually by the BeefTrader platform (FAPESP 2015/07855-7) and LPT (2018/01184-1), with 1,000 animals evaluated, improving the database with data of individual. The main products are: i) a camera system will be developed to monitor animal behavior, with static position and attached to the feed mixer; (ii) a database will be created to associate: animal information, consumption, orts, climatic data and image of the animal attitude and trough conditions; iii) will be developed classification models based on deep convolutional networks to classify the animal images and the trough score; iv) will be developed an algorithm for predicting animal consumption, considering animal attitude, diet data, consumption, orts and climatic events; v) will be collected information and requirements to create a knowledge through a report of automation feed mixer using Internet of Things (IoT); and vi) to present a international congress manuscript. (AU)