New Machine Learning-Based Tool Uses Sketches to Improve Tumor Detection

Researchers at the University of Surrey have developed a unique sketch-based object detection tool that leverages AI to search within an image for specific objects. By sketching an object, the AI uses visual cues such as pose and structure to find a matching object while disregarding more generalized options. According to Professor Yi-Zhe Song, the tool could revolutionize cancer detection and wildlife conservation by matching specific objects and discounting others. The technology will be presented at the upcoming Computer Vision, Pattern, and Recognition Conference (CVPR). The researchers use an example of the tool searching for a single zebra, using only a sketch of a zebra eating in a picture full of zebras. The AI tool bases its decisions off the exact requirements given by the artist, allowing humans to interact with AI from a different perspective. This translates to more human creativity in computer vision and keeps necessary human intervention.