Procyon® AI Computer Vision Benchmark 2.0
The Procyon AI Computer Vision benchmark provides you with valuable insights on how AI Computer vision tasks perform on your Windows Pc or Apple Mac. The workload has been designed around a range of machine vision tasks and AI models, carefully selected based on their use case, relevance and daily impact in the modern office.
Features
- Featuring support for the new Windows ML standardized Execution Provider (EP), existing support remains for native inference engines NVIDIA® TensorRT™, Intel®OpenVINO™, Qualcomm® SNPE and Apple® Core ML™.
- Developed in partnership with AI leaders with machine-vision tasks in mind.
- Measure inference performance using the GPU or NPU.
AI Models
During development of the latest 2.0 version of the Procyon Computer Vision Benchmark, we carefully selected machine vision tasks and artificial intelligence (AI) models based on use case, relevance and daily impact in the modern office.
Image classification - ConvNeXt-Tiny
In AI image classification, the AI model inspects an image or video frame and classifies its contents. Some uses for image classification include smart visual search functions such as searching, sorting and tagging images or videos in a content library, and retail inventory management.
Image captioning - BLIP (Base)
Image captioning refers to generating natural‑language descriptions of an image using an AI model that combines visual understanding with language generation. This workload mirrors several emerging Windows 11 scenarios, such as AI‑enhanced accessibility capabilities, smart content tagging and visual summary features in productivity tools.
Video object detection - Base DETR
Object detection states what is in the image and where each object is, for example, the width, height and class label. It is a key tool for differentiation, and anything that needs identification and localization uses object detection.
Video segmentation - SAM2
Video or image segmentation is the technique of identifying and partitioning regions of an image or video frame into regions. AI image or video segmentation is used for tasks such as blurring the background of a video or applying masks to objects.
Video upscaling - Real-ESRGAN
AI-enhanced upscaling takes an initial video or image and improves its fidelity by using AI to work out and re-add missing information. This AI use case can be used to improve low-quality images or video feeds, or reduce the bandwidth needed for a clear picture, such as for a video call made from a place with poor signal.
Read more details about Computer Vision Benchmark 2.0 in our user guides.
Reduce complexity, make sense of AI scores and monitor your hardware in one place
- Condense complex spec sheets, software versions, power saving states and other hidden details that affect performance into an easy-to-understand and record single score.
- Test with a GPU or NPU and get further scores showing inference times and more for each neural network test.
- Get detailed metrics on how GPU temperatures, clock speeds and component usage change during the benchmark run.
Computer Vision 1.0
The first version of the Computer Vision benchmark launched in 2023 and utilized the MobileNet V3, Inception V4, YoloV3, DeepLab V3, Real-ERSGAN,and ResNet 50 AI models.
It’s now been succeeded by our latest 2.0 version, but you can still access it in Procyon and is still supported as some of the models still have use cases.
Read more about Computer Vision Benchmark 1.0 in our user guides.
Site license
get Quote Press license Request trial- Annual site license for Procyon AI Computer Vision Benchmark.
- Unlimited number of users.
- Unlimited number of devices.
- Priority support via email and telephone.
Benchmark Development Program
Contact us Find out moreThe Benchmark Development Program™ is an initiative from UL Solutions for building partnerships with technology companies.
OEMs, ODMs, component manufacturers and their suppliers are invited to join us in developing new AI processing benchmarks. Please contact us for details.
Minimum system requirements for Windows PC
| OS | Windows 10, 64-bit or Windows 11 |
|---|---|
| Hardware | Supported AI accelerator |
Minimum system requirements for Apple Mac
| OS | macOS Tahoe 26.0 or later |
|---|---|
| Processor | Apple Silicon |
AI Quality Metrics
While comparing inference engines, it’s important to consider accuracy in addition to raw performance.
We’ve run our own tests measuring the accuracy of inference engines supported by the Procyon AI Inference Benchmarks.
Support
Latest 1.0.539 | Mar 30, 2026
Languages
- English
- German
- Japanese
- Portuguese (Brazil)
- Simplified Chinese
- Spanish