# YunKan-CUDA

YunKan-CUDA is a self-hosted, local-first NVR / video management system (VMS) for home and small-business IP cameras that uses an NVIDIA GPU to run all AI inference on your own hardware. It is open source and self-hosted, installable on Docker Hub and Unraid.

Canonical page: https://ourlocalhost.com/app/2291/yunkan-cuda

## Description

YunKan-CUDA is a self-hosted, local-first NVR / video management system (VMS) for home and small-business IP cameras that uses an NVIDIA GPU to run all AI inference on your own hardware. It detects motion, people, vehicles, and faces, reads license plates (ALPR / LPR), and recognises poses, falls, packages, and baby cries — with all footage kept on your LAN instead of a vendor cloud. The full stack ships pre-bundled in a single container, so there is no separate message broker, database, reverse proxy, or model download to configure by hand. It is the CUDA build of the four-variant YunKan family (CPU, CUDA, TensorRT, OpenVINO; only one variant can run at a time) and a privacy-first alternative to Frigate, Shinobi, ZoneMinder, Scrypted, Agent DVR, iSpy, and MotionEye. Recommended for homelabbers who already run an NVIDIA GPU in their Unraid server, with at least 6 GB of video memory (VRAM; 8 GB recommended for many high-resolution cameras).

## Platform availability

Available on 1 of 10 platforms: Unraid.

## Health score

50/100 (Fair), based on 4 of 4 factors.
- Maintained: 100 (last commit 25 days ago)
- Popular: 6 (1 GitHub stars)
- Easy to install: 35 (1 install platform, install notes, Docker image)
- Light to run: 10 (6 GB minimum RAM)

## Links

- Docker Hub: https://hub.docker.com/r/martin888/yunkan-cuda
- GitHub: https://github.com/mrtian2016/yunkan-unraid-templates
- Unraid: https://ca.unraid.net/apps/yunkan-cuda-0qczhoq0g80fo6?q=YunKan-CUDA
- Official website: https://yun-kan.com

## Common questions

### Can I use YunKan-CUDA with an older GPU without TensorRT engine-build failures?

Yes. If TensorRT cannot target an older GPU or driver, the unRAID template points users to YunKan-CUDA, which skips engine building and starts without waiting; CUDA is the safe default and TensorRT is opt-in.

### Is YunKan-CUDA charged per camera?

No. It is licensed per household rather than per camera.

### Are native mobile apps available for YunKan-CUDA?

Yes. It includes native Android and iOS apps along with a web interface.

### Does YunKan-CUDA upload video recordings to a cloud service?

No. It runs fully local and self-hosted, so footage stays in the home and does not pass through vendor servers.

---
Source: Our Local Host (https://ourlocalhost.com/app/2291/yunkan-cuda), the self-hosted and open-source app directory.
