Instructions | MNIST on ESP32-C3 — No TensorFlow, Pure C

Clone the repository

git clone https://github.com/AlexRosito67/hasaki-mnist-esp32

2

Open in Arduino IDE or PlatformIO

Open `src/main.cpp` (PlatformIO) or `mnist_esp32.ino` (Arduino IDE).

Set your WiFi credentials

#define WIFI_SSID     "YOUR_SSID"
#define WIFI_PASSWORD "YOUR_PASSWORD"
#define DISPLAY_IP    "192.168.1.XXX"  // optional LED matrix

4

Select board

ESP32C3 Dev Module (Arduino IDE) or `esp32-c3-devkitm-1` (PlatformIO).
5

Flash

Upload to your ESP32-C3.
6

Get the IP

Open Serial Monitor at **115200 baud**. The ESP32 prints its IP on startup.
7

Draw a digit

Open `http://<ESP32_IP>` on any browser on the same network. Draw a digit and click **Send to ESP32**.
8

Note on the model header

The repository includes `mnist_int8.h` — a pre-trained INT8 model ready to use.

To train your own model with different architecture or parameters, use Hasaki
https://github.com/AlexRosito67/hasaki.

MNIST on ESP32-C3 — No TensorFlow, Pure C