ADI EdgeBench software showing audio denoiser demo and inference results.

Overview

Features and Benefits

  • Run models on real ADI hardware remotely. No lab access needed. Execute inference directly on physical evaluation boards from anywhere.
  • Zero-setup, zero-touch workflow. No firmware flashing, SDK installation, or board bring-up. Login to get results in minutes.
  • Hardware-accurate inference metrics. Real runtime latency, memory footprint, and on-device behavior, not simulated approximations.
  • Curated model zoo. Benchmark pre-validated, device-optimized architectures ready to run out of the box.
  • Upload your own models. Evaluate TensorFlow, PyTorch, and ONNX models with documented packaging rules.
  • Reproducible benchmarking. Every test runs in a controlled, standardized environment for reliable, comparable results.
  • Secure, 24/7 board access. Scalable back end handles concurrency, provisioning, and session management automatically.
  • One workspace for everything. Models, datasets, builds, and multitarget projects all in one place.
  • Unified, browser-based UI. Hardware selection, model configuration, and result analysis—no tool-switching.
  • Accelerate hardware selection. Validate device fit early, compressing evaluation cycles from weeks to days.

Compatible Parts

How it Works

EdgeBench streamlines early AI model evaluation by giving engineers instant, browser‑based access to real ADI hardware through the Remote Board Farm. The platform handles compilation, device setup, firmware flashing, and inference automatically, eliminating hardware bring-up and toolchain work. Engineers can test ADI Model Zoo networks or upload TensorFlow, PyTorch, and ONNX models to assess feasibility. EdgeBench provides hardware‑accurate latency, memory, and behavior metrics, enabling confident device selection and faster iteration. With consistent lab conditions and 24/7 secure access, teams reduce evaluation delays and derisk edge‑AI development.


Project and Model Management

Seamlessly manage projects and uploaded models through an organized, easy‑to‑navigate workspace dedicated to storing builds and BYOM assets.

ADI EdgeBench UI showing a list of AI/ML projects.


Device Configuration, Compilation, and Deployment

Device configuration, compilation, and deployment are executed through UI-driven workflows, eliminating manual build setup, SDK installations, and direct flashing or debugging steps.

EdgeBench UI: MAX32690 Device Selected


Results Visualization

View inference outputs and performance metrics through interactive visualizations.

Anomaly detection inference results showing input and MSE loss.


ADI EdgeBench

EdgeBench is designed to be the trusted starting point for AI evaluation on ADI hardware, helping teams move from idea to validation with clarity, speed, and accuracy.

Systems Requirements

  • Web-based application (no installation required)
  • Free with ADI account login

Documentation & Resources

  • View All (2)
  • User Guide (1)
  • Frequently Asked Question (1)
    • Document Type