Cells Counting with Convolutional Neural Network

In this paper, we focus on the problem of cells objects counting. We propose a novel deep learning framework for small object counting named Unite CNN (U-CNN). The U-CNN is used as a regression model to learn the characteristics of input patches. The result of our model output is the density map. Density map can get the exact count of cells, and we can see the location of cell distribution. The regression network predicts a count of the objects that exit inside this frame. Unite CNN learns a multiscale non-linear regression model which uses a pyramid of image patches extracted at multiple scales to perform the final density prediction. We use three different cell counting benchmarks (MAE, MSE, GAME). Our method is tested on the cell pictures under microscope and shown to outperform the state of the art methods.