Chooch is a horizontal computer vision platform designed to train visual AI models according to bespoke use case requirements. Its unique value proposition is driven by four primary capabilities: rapid dataset generation through various techniques (such as synthetic data generation, video annotation, and smart annotation features), model training & hosting (created with Chooch, uploaded from external sources, and exportable), identifying both objects and actions, and programmatic edge device deployment as an inherent platform offering. These core competencies allow Chooch to compress the machine learning timeline for data annotation, training, and model deployment by an order of magnitude faster than competing platforms. Additionally, Chooch is designed for edge deployment on Intel CPU and NVIDIA GPU hardware, which enables Chooch to immediately deploy edge inferencing into disconnected environments and capture insights every 16 to 20 milliseconds on average.

The inference engine can deploy to the edge on 9th generation Intel Core CPUs i7 and later. On a 9th generation Intel i7 CPU, the platform can perform inferencing on up to 4 camera streams with 2 models each or 8 cameras with 1 model each. On NVIDIA Jetsons, Chooch AI can support 3 to 4 camera streams with single models, or a single camera with 3 to 4 models total. On NVIDIA Tesla T4 GPUs, Chooch AI can support 15 to 20 parallel streams. Similarly, 15 to 20 cameras can perform inferencing with a single model, or 3 to 4 cameras can run 4 to 5 models total.

The entire platform - data preparation tools, model training & hosting, inferencing, and analytics - is available in both cloud and self-hosted environments.