Hands-On Machine Learning for Algorithmic Trading

Computer vision beyond classification – detection and segmentation

Image classification is a fundamental computer vision task that requires labeling an image based on certain objects it contains. Many practical applications, including investment and trading strategies, require additional information.

The object detection tasks require not only the identification, but also the spatial location of all objects of interest, typically using bounding boxes. Several algorithms have been developed to overcome the inefficiency of brute-force, sliding-window approaches, including region proposal methods (R-CNN) and the You Only Look Once (YOLO) real-time object detection algorithm (see references on GitHub).

The object segmentation task goes a step ...