deployment-with-CMSIS-NN

1. check out official tutorial and documentations
2. check out Background_Knowledge.ipynb
3. check out Quantization.ipynb
4. run CMSIS_NN_PC_simulator in Visual Studio to verify the results
5. deploy the model on Cortex M4 boards

Per-tensor quantization in CMSIS NN (official tutorial)

https://community.arm.com/developer/ip-products/processors/b/processors-ip-blog/posts/deploying-convolutional-neural-network-on-cortex-m-with-cmsis-nn

https://developer.arm.com/solutions/machine-learning-on-arm/developer-material/how-to-guides/converting-a-neural-network-for-arm-cortex-m-with-cmsis-nn

Documentation

https://www.keil.com/pack/doc/CMSIS/NN/html/index.html

https://github.com/ARM-software/CMSIS_5

Problems

1. Quantization range is not fully covered. (we don't need to represent the negative value before ReLU)
2. 32bit overflow. ('16 bit weight * 16 bit tensor + 16 bit bias' can exceed the limit of int 32)
3. how to generate the code automatically (https://github.com/majianjia/nnom)

STFT

ARM_STFT_ISTFT.c: STFT and ISTFT for ARM boards (same with libosa's implementation)

Cite

@article{lee2024optimizing,
  title={Optimizing RGB Convolution on Cortex-M7 MCU: Approaches for Performance Improvement},
  author={Lee, Seon-Woo and Sung, Jae-Mo and Kwon, Jang-Woo},
  journal={IEEE Access},
  year={2024},
  publisher={IEEE}
}