Skip to content

README.md

Latest commit

 

History

History
127 lines (85 loc) · 4.21 KB

README.md

File metadata and controls

127 lines (85 loc) · 4.21 KB

Geti Instant Learn Application

Full-stack web application for deploying zero-shot visual prompting models on live video streams, cameras, and video files.

Built with the Geti Instant Learn Library for model inference, FastAPI for the backend, and React for the frontend.

License acceptance

By installing, using, or distributing this library/application, you acknowledge that:

  • you have read and understood the license terms at the links below;
  • you confirmed the linked terms govern the contents you seek to access and use;
  • you accepted and agreed to the linked license terms.

License links:

In order to consent, set an environment variable INSTANTLEARN_LICENSE_ACCEPTED=1 or accept terms when first importing the library.

Quick Start

Run from Source (Development)

Prerequisites: uv, Just, Python 3.12+, Node.js v24+

# Start backend and frontend in development mode
just device=xpu application/dev

Access at: http://localhost:3000

Configuration parameters
Variable Default Description
port 9100 Backend API port
ui-port 3000 UI development server port
device cpu Hardware target (cpu, xpu, cu128)
enable-coturn false Enable local TURN server for WebRTC
stun-server "" External STUN server URL
coturn-port 443 Port for local TURN server

Note: WebRTC parameters configure video streaming between browser and backend. See WebRTC Networking for deployment scenarios.

Run with Docker

Prerequisites: Just, Docker

Build the image:

# Build for Intel XPU (recommended)
just device=xpu application/build-image
Build parameters
Variable Default Description
device cpu Hardware target: cpu, xpu, cu128
build-target cpu Docker build stage: cpu, xpu, cuda
version latest Image version tag

Run the image:

# Run with default settings
just device=xpu application/run-image

Access at: http://localhost:9100

Runtime parameters

Networking:

Variable Default Description
port 9100 Port for serving UI and API
webrtc-ports 50000-51000 UDP port range for WebRTC
stun-server "" External STUN server URL
enable-coturn false Enable local TURN server
coturn-port 443 Port for TURN server

Note: WebRTC parameters configure video streaming between browser and backend. See WebRTC Networking for deployment scenarios.

Hardware:

Variable Default Description
device cpu Hardware target: cpu, xpu, cu128
webcam-device /dev/video0 Path to webcam device

Documentation

Getting Started:

Using Geti Instant Learn:

Concepts: