ESP32_NN_by_hand

This project is about how to run the forward process of a Nerual Network on an embedded system.

Environment

Programming Environment: ESP-IDF (v5.1.1)

Microprocessor: ESP32-S3 N16R8

Function

1. conv1d layer
2. flatten layer
3. linear layer
4. ReLU layer

Example

This example donates a simple neural network, including a conv1d layer and a linear layer.

The size of input, output, weights and bias and the model itself are in model.h.

The hello_world_main.c file defines the neural network and uses a dummy input data to test the network.

You can modify these two files to define and run your own network.

Get Start

Make sure you have downloaded the ESP-IDF and Git before getting start!

1. download the code

git clone https://github.com/wjc1207/ESP32_NN_by_hand.git

2. clean the build

idf.py fullclean

3. build and flash the code to the ESP32

idf.py flash -p <port>

4. open the monitor to see the results

idf.py monitor -p <port>

If everything works well, you should see the following output:

0.046326
0.121813
0.021277
-0.025692
Time: 8176

Tools

The convert_json_to_h.py can be used to convert the model in json file into the .h file directly.