A trainable layer that switches how ML blocks operate on images based on the contents of those images at the pixel level. - View it on GitHub