Counting Manatee Aggregations Using Deep Neural Networks And Anisotropic Gaussian Kernel

This article discusses a new approach for manatee counting using Anisographic Gaussian Kernel (AGK) based crowd counting. The proposed method employs line-segment annotation, using a single line-segment to label each manatee. The input of the framework is an image, represented as a matrix. The deep neural network is trained to predict the density map for an image, such that the network predicted density map output is sufficiently close to the ground-truth. A kernel, commonly used in machine learning to perform classification and clustering, is a non-linear mapping of two vectors in a feature space, through the dot product of two vectors.