Neural Network for Spectrum

Introduction

This is a Project for people who doesn't know Neural-Network-Programming to train their data with simply click buttons on the window. To get the right result, you must strictly respect the following Data-Format:

1. Train Dataset (File format: .csv)

Train Dataset should be csv file with the first column is class, the rest columns are feature values. One row is one sample. It should look like this:

2. Test Dataset (File format: .csv)

Test Dataset also should be csv file. All columns are features values.

3. Test Image (File format: .csv)

Test Image also should be csv file. If the size of Image is (320, 20), then the length of test file should conclude: 320 x 20 = 6400 rows.

Usage

Install the required packages with following command.

pip install -r requirements.txt

Start the Application

python Start.py

Then you will see this window:

Train

Press the Choose Train File to choose your train_set file, you can choose open Data Distribution Function or not. Data Distribution Function may take a while to draw the result picture, you need to wait until the result image shows.

Press Start Train to start the training task, set the test_size and epochs, then you can see the progress in the terminal window.

Press Save Model to save this model.

Test

To test your model, you need to load model at first. There are two modes to do prediction (test row data & test image).

1. test row data

Choose your row data file, you can get this row's material class in the terminal like this:

1. test Image data

Choose your row data file, you will see the color image which has already mapped the material ID to the specific color like this: