Density Map Estimation for Crowded Chicken

摘要

Intensive breeding is the trend of the breeding industry. In order to make it more convenient to manage and reduce labor costs, sometimes we need to estimate the number of individuals in the poultry farm and discriminate the density distribution to help scientific management. At the same time, crowd density estimation is a developing research direction in deep learning. There are both similarities and differences between crowd counting task and chicken counting task. Aimed at the characteristics of poultry farm images, this paper presents a solution to density estimation and counting of poultry individuals in poultry farm by deep network method. We designed an end to end model and transform the problem into a pixel-level classification problem to get the density map.