FIG. 10 is a block diagram illustrating an example of a machine learning model. In this example, model 1000 employs a convolutional neural network (CNN) to generate a feature map based on input channels of an image, and can be used to implement 204 of FIG. 2.

The CNN is a type of deep learning neural network for analyzing images and identifying features. Any appropriate CNN implementation can be used, such as Faster RCNN, SSD, YOLO, etc. In this example, 1002-1006 are the Red(R), Green(G), and Blue(B) channels, respectively. A three-dimensional matrix is used to represent the channels (with dimensions X and Y corresponding to height and width of the images, and dimension Z corresponding to the channels). The matrix is sent to the CNN as input. The CNN includes multiple layers, where the first layer applies a convolutional filter to the input and each subsequent layer applies a different convolutional filter to the output of the previous layer. The successive layers each detect a specific type of data (usually a higher level of feature than the previous layer). For example, the first CNN layer detects edges in horizontal, vertical, or diagonal directions, the second CNN layer detects curves based on the previously detected edge data, and the third layer detects features, etc. Additional layers can be used.