Aidge: Independent Deep Learning framework for embedded AI

Aidge is an open source and collaborative framework dedicated to embedded AI that offers a complete tool chain to design, optimize and deploy deep neural networks for constrained systems. Aidge integrates innovative features to reduce the complexity of models and their memory needs and enables optimized implementation on a wide range of hardware architectures such as MCU, CPU, GPU, FPGA, and ASIC.


  • Aidge is hosted by Eclipse Foundation for open and business friendly environment
  • Aidge takes benefit of strong French and European initiatives (DeepGreen and NeuroKit2E projects) to become the sovereign reference platform for embedded AI



European countries


main projects

Innovation & assets

  • Support of the State of the Art Models: CNN, RNN, GAN and Transformers architectures
  • Powerful graph manipulation: Tiling, Graph search and replace Engine
  • Post training Optimization: Quantization, Pruning, Compression
  • Quantization Aware Training: Innovative and state of the art method based on SAT and LSQ
  • Robust approaches for learning and inference: Incremental learning and Adversarial attack
  • Export modules for optimized and transparent code generation on a wide range of hardware targets: MCU, CPU, GPU, NPU, ASIC, FPGA
  • Hardware Design: NeuroCorgi
Use Cases

Uses cases on key sectors

  • Space & Defense: On-board recognition and geolocation for drones and satellites, Operator assistance functions using on-board AI in the defence context
  • Automotive: object detection in urban environment in the ADAS, Advanced Driver Assistance Systems context
  • Avionics: functions development for safe and trusted avionics
  • Smart Manufacturing: Image recognition for production control in constrained environment, Predictive maintenance for autonomous systems
  • Health: real time speech enhancement on a miniaturized solution integrated on glasses for the hearing impaired

Our offer

  • Free access to the tool
  • Specific module development on demand
  • Expertise on embedded AI to reach the requested performance in terms of accuracy, latency, power consumption, form factor…


To go further:


  • Sandrine Varenne
    Partnership Manager