EdgeFirst has published the EdgeFirst Model Zoo on Hugging Face, featuring YOLOv5, YOLOv8, YOLO11, and YOLO26 models validated across 20 distinct edge platforms. The validation relies on a single Rust pipeline that handles zero-copy DMA pre-processing, model decoding, and post-processing directly on the target hardware.

  • Platforms include NXP i.MX 8M Plus, i.MX 95, Ara240, Raspberry Pi 5 with Hailo-8L, Jetson Orin Nano, Qualcomm Hexagon, Apple Neural Engine, and x86/Arm/CUDA references.
  • The pipeline runs the full COCO val2017 dataset on-device, generating predictions.parquet, trace.pftrace, and metrics.yaml for each of the 837 sessions.
  • Measured throughput accounts for stage overlap across frames, revealing bottlenecks such as post-processing limits rather than NPU constraints.

This approach allows users to recompute metrics from the Parquet files alone and inspect detailed pipeline traces to understand performance characteristics on specific edge devices.