CUDA Vision & Imaging Library

Simple to use, GPU accelerated computer vision SDK.

Check latest release notes.

Build realtime and robust vision apps with CUVI

Start building your own computer vision solutions using CUVI

Ease of Use

Requires less than 5 minutes to setup and start using CUVI on your machine. The framework comes with intuitive interface and requires little programming experience from user.

Exceptional performance

Perform at 100 plus FPS on 4k images for most vision pipelines. In comparison, CUVI performs magnitudes faster than Intel® IPP™, OpenCV and other vision libraries.

Feature Rich

Provides 100 plus high-end image processing primitives and variants ranging from medical, industrial and defense domains. These features are all handcrafted and highly optimized for GPUs.


CUVI fuels many real world applications that are running 24/7 without failure ranging from underwater to land to drones. All features are robust, highly tested and memory friendly.

Top news and articles from CUVI's blog.

We write about exciting news in the GPU industry and also latest releases and guides from CUVI library,
its performance benchmarks and comparison with other SDKs.

cuvi nvidia

CUVI becomes NVIDIA Partner Library

CUVIlib is now officially part of NVIDIA CUDA-X and is listed under partner libraries on NVIDIA’s website. Our journey began in early 2010 when we envisioned a library built on top of CUDA to provide accelerated primitives for vision applications. Since then we have helped dozens of companies and thousands of users with their performance needs. Today, we are better, faster and more robust than ever. [...]

cuvi underwater

Aerial Object Tracking on CUDA

Unmanned aerial vehicles or UAVs with a real-time video feed offer numerous vision applications. Many of these applications depend on detection of objects and tracking those objects along the scenes. The feedback from tracking data can also be used to steer UAVs to follow that object or person of interest. In this article we’ll use code example to show how this can be done using CUVI. [...]

cuvi dfpd

Robust DFPD debayer algorithm now in CUVI

Continuing the legacy to provide the best imaging algorithm at lightning fast speed, we are proud to announce the addition of DFPD debayer algorithm in CUVI which is more robust than the existing demosaic and shows no artifacts at high feature areas. Linear demosaic implementation though is super fast giving a throughput of more than 500 fps on HD image on a common GPU yet it has its downside. [...]

Care to read more? Visit our GPU Blog.

Download CUVI.

Supports Win (64-bit) and Linux (64-bit & aarch64)

You'll be asked to sign up for CUVI updates after which you will receive download links. For information on the functionality available
with this download, refer to the Wiki page. The free version is limited to 1000 x 1000 image size.

Listen to our happy users

Jean-Pierre Gehrig, Co Founder Edeltech

Thanks to CUVI we added real time 4k image processing to our film scanning application without writing one single line of GPU code. Also, we reduced our hardware requirements from a 12-core Xeon server to a single Ge-Force card.

Remco Schoenmakers, Director Thermo Fisher Scientific

CUVI allowed us to keep the processing power in line with ever increasing data volumes of modern detectors without sacrificing user experience for our electron tomography applications.

Ian Holmes, Founding Director underseaROV

CUVI helped us put underwater enhancement, a rather compute intensive pipeline, into mobile rovers. The results are impressive and we are able to achieve real time performance on full HD video via a laptop GPU.

Need help or have questions?

Contact us for any help regarding CUVI, off-the-shelf vision solutions, GPU based product development and code porting with performance analysis.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.