The UL Procyon AI Computer Vision Benchmark performs inferences using each model for 3 minutes and records the average inference time, median inference time and the total number of inferences performed.
The benchmark allows the user to select the inference engine, target hardware and model precision. Once the benchmark completes, the overall score is calculated by taking into account the average inference times of the models and a scaling constant used to bring the score in line with the traditional range for UL benchmarks. The higher the score, the better the performance.
AI Inference score = K * (1 / geometric mean of MobileNetV3 Average Inference Time, ResNet50 Average Inference Time, InceptionV4 Average Inference Time, DeepLabV3 Average Inference Time, YOLOV3 Average Inference Time, Real-ESRGAN Average Inference Time) Where K = 5000
K is a scaling constant used to bring the score in line with the traditional range for UL benchmarks.
The benchmark result details screen displays the recorded average inference time, median inference time and the total number of inferences for each model.