Sliding Window Machine Learning Approach For a single object detection task, the idea is to train a binary classifier, which determines if the presented object is “positive” or “negative.” The trained classifier can then be used to “inspect” a target image by sampling it, starting from the top-left corner.