Algorithm 1: HOG feature extraction.

Input: Image dataset with RGB colors. Step 1: Compute the gradient of each pixel of the image, number of orientation bins used was 9. Step 2: Divide the images into cells, 8 × 8 pixels form a cell, and compute gradient histograms of each cell. Step 3: 2 × 2 cells form a block, and normalize gradient histograms across blocks. Step 4: The feature descriptors of all blocks are then flattened into a feature vector. Output: HOG features.