The Procyon AI benchmark measures inference times on popular models like MobileNet, ResNet, and YOLO. It's designed to show how well a system can accelerate machine learning tasks in real use.
In UL Procyon AI Computer Vision, the K16 scores 69 overall, with 1.86 ms in MobileNet V3, 18.58 ms in ResNet 50, 46.32 ms in Inception V4, 84.05 ms in DeepLab V3, 122.22 ms in YOLO V3, and 8608.79 ms in REAL-ESRGAN. The overall result is low, which fits the broader profile of the system, but a few of the lighter model results are better than I expected. Even so, this is not one of the K16's stronger areas, and the heavier models show the limits of the hardware pretty quickly.