Machine vision
Explore dataset creation resources, pre-trained models, and full AI pipelines.
Every pipeline has latency — capture, inference, post processing. Target velocity is measured frame to frame and the tracking point is projected forward, so your crosshair stays on target rather than where it used to be.
Detection boxes aren't always accurate. Certain poses/animations [I.e gliding] may make the detection box taller than the hitbox. In this scenario, a static aim height would cause you to shoot above the target. Smart aim detects when the target is horizontal and aims lower on the detection box so you don't miss.
Detection models have an optimal input size. If the target is far away, the model may lose the detection. Dynamic scaling will upscale small targets and downscale large targets to keep the optimal target size.
Output runs over dedicated microcontrollers, isolated from the host machine. Firmware ships prebuilt for four boards and flashes in one step.
Bring Your Own Model. We do not lock down our models. You can use any YoloV5 through Yolo26 model. Models can be found in the community tab of the account page, or at mossymodels.com