STANN

STANN (Synthesis Templates for Artificial Neural Networks) is a template library used to build FPGA-based neural network accelerators via High-Level Synthesis.

This project cannot be synthesized on its own but can be used by other projects to implement neural networks on FPGAs.

Example project

The following project demonstrates how to use STANN to implement LeNet-5 and TenNet (simple neural network with 10 layers): https://github.com/ce-uos/stann-example

Citation

If you use this repository, please cite this paper:

@InProceedings{stann2023,
    author="Rothmann, Marc and Porrmann, Mario",
    title="STANN - Synthesis Templates for Artificial Neural Network Inference and Training",
    booktitle="17th International Work-Conference on Artificial Neural Networks (IWANN2023)",
    year="2023",
    publisher="Springer International Publishing"
}