Test accuracy of imagenet data set on MCU？

mit-han-lab / tinyengine

[NeurIPS 2020] MCUNet: Tiny Deep Learning on IoT Devices; [NeurIPS 2021] MCUNetV2: Memory-Efficient Patch-based Inference for Tiny Deep Learning; [NeurIPS 2022] MCUNetV3: On-Device Training Under 256KB Memory

https://mcunet.mit.edu

MIT License

757 stars 127 forks source link

Test accuracy of imagenet data set on MCU？ #84

Open ahdxwg opened 1 year ago

ahdxwg commented 1 year ago

Great work, I was wondering how you tested the accuracy of testing the imagenet data set on MCU? Is it actually running on MCU? How is the test data stored on the MCU? I would like to do a baseline thank you very much.

meenchen commented 1 year ago

Hi @ahdxwg,

We simulate the on-device inference with PyTorch (similar to how you do quantization-aware training) and measure the accuracy on servers.