Software
Services
Support
Insights
Compare
More
While comparing inference engines in computer vision workloads, it’s also important to consider accuracy in addition to raw performance. Converting the different AI models between formats and quantizing to different precisions to run using different inference engines can affect the quality of object recognition in a given image.
To support comparisons between platforms, we’ve run our own tests measuring the accuracy of inference engines supported by the Procyon AI Computer Vision Benchmark.
The following interactive graph shows the computer vision models & inference engines tested by UL Solutions. Along the X-axis are the accuracies of the tested models, while along the y-axis you can find the AI engines and devices. The data is grouped according to the quality metric specific to the ai use case and precision of the model.
This was tested by UL Benchmarks for the models used in AI Computer Vision Benchmark (v1.5.290 for Windows and v1.0.58 for macOS) and the real world quality of the models may vary depending on the source of the model and test data set.
Flexible licensing lets you pick and choose the individual benchmarks that best meet your needs. Buy just one benchmark or add more in any combination.
UL Procyon is a new benchmark suite from UL that we’re creating specifically for professional users in industry, enterprise, government, retail and press. Each Procyon benchmark will provide a familiar, consistent experience by sharing a common design and feature set. Procyon benchmarks measure performance in a variety of real-world use cases. It offers benchmarks for AI Inference, Office Productivity, Battery Life, Photo Editing, and Video Editing.