[ML Search] Major performance improvements - in most cases you can expect processing performance at about 60FPS even for large (1024x1024+) models. This opens up a whole new world for automation.
[ML Search] ML now provides a bindable position(Bounds, BoundsWindow and BoundsScreen) of the best match, which means it's possible to configure SendInput without writing a single line of code. Note than “Link aura” is not supported yet for ML, only bindings.
[ML Search] Added YoloV8 Classification support - basically these models provide a probability that whatever is on an image is of specific class(dog, cat, etc) without providing more details
[ML Search] Added YoloV8 Segmentation support - extremely powerful detection type which provides not only bounding boxes of found object, but also a mask which represent an exact form of an object. This is useful only for scripting purposes atm.
[ML Search] Fixed dozens of issues with Machine-learning-based Search
[Custom UI] A bunch of improvements - fixed few crashes, improved editor autocompletion, etc
[Overlays] Fixed deadlock that occurred when there were multiple bindings which change position of overlay at the same time (e.g. size+offset size)