Tasks
Pick by the output you need. Every model of a task ran on the same shared inputs on our RTX 4090. Tasks are added as models pass the protocol.
Vision5 tasks
- object-detection
Object detection11 modelsFind objects and draw a labeled box around each one.Real-time, closed set, Open vocabulary - instance-segmentation
Instance segmentation6 modelsA separate pixel mask for every object.General objects - image-matting
Image matting2 modelsA soft alpha matte that separates the subject from the background, hair included.Background removal - depth-estimation
Depth estimation4 modelsHow far away each pixel is, from a single photo.Monocular, relative, Monocular, metric - pose-estimation
Pose estimation5 modelsBody keypoints and skeletons for every person.Multi-person 2D
