Dynamic Visual Data Segmentation and Discrete Tracking

Alaya AI’s image segmentation and tracking system is developed based on OpenCV’s Computer Vision Annotation Tool and an enhanced version of Meta’s Segment and Track Anything. This system supports both static image processing and dynamic video analysis, enabling efficient and seamless instance segmentation and tracking while minimising manual input workload for multi-frame labelling and continuous object tracking.

Alaya AI’s auto-labelling system is capable of automatically identifying regions-of-interest (ROIs) in dynamic visual environments and automatically identify segmentation elements and scenes. If the segmented scene is recorded in a continuous video, the system will be able to apply AI algorithms to maintain the tracking and association of the ROIs automatically without extensive human intervention.

Hardware Specifications

The implementation of the Alaya system will include a backend AI server and a frontend UI web server. The backend AI server can be packaged into docker to be deployed in one or many GPU servers concurrently. To avoid latency in the backend computation, a backend AI server should have at least one Nvidia 3090 or 4090 GPU, latest Intel or AMD gaming-grade CPU, and at least 32GB of main memory.