Automatic Fecal Eggs Counting in Ruminants Using Xilinx DPU

Abstract

Parasite control is vital in the cattle industry. Infection control is based on counting parasite eggs on samples obtained from animals, and processed later in laboratories by human technicians using optical microscopes. This traditional monitoring method affects the yield of production, delaying the start of treatment of infected animals. It also implies an additional cost of mobility of the veterinary expert who determines the treatment. Therefore, it is convenient to have a portable device for automatic fecal egg counting. This work presents an embedded algorithm for this type of analysis, with the ability to implement a portable solution. The implementation is based on Deep Learning, running on a MPSoC (Multiprocessor System-on-Chip) using Xillinx DPU. Our proposal has resulted in significantly lower sample processing time compared to other existing solutions, greater flexibility and a low power consumption.