Orienfish / arduino-dataloader

arduino-dataloader

Setup a SD card, store the ML dataset in the SD card, and load the dataset on Arduino board.

Getting Started

This quick setup takes Arduino Nano BLE 33 and MNIST dataset as an example.

• Connect the SD card reader with Arduino (using SPI). Format the SD card to FAT32.

• Download and unzip the MNIST dataset. Copy the prepared MNIST dataset (in ./laptop/MNIST) to the SD card.

  Note, that due to the limitation on the filename length on the SD card, I rename the downloaded datasets:

  train-images-idx3-ubyte    -> tr-img // Training images
  train-labels-idx1-ubyte    -> tr-la  // Training labels
  t10k-images-idx3-ubyte     -> te-img // Testing images 
  t10k-labels-idx1-ubyte     -> te-la  // Testing labels
  	
  

• Use load_data/load_data.ino to load the dataset. Insert your code for training here.