Skip to content

How to use AutoNN GUI for Image Dataset

The GUI interface

Steps to run GUI :

  • Open terminal
  • Run the following command 
    pip install nocode-autonn 
autonn

The GUI interface

Buttons

  • Open folder : Used to select the path to the training dataset
  • Show Configs : Displays all the initial settings before training process
  • Predict : To make predictions on selected images
  • Load Model : To load a trained model
  • Display Graphs : Displays the Training loss/accuracy vs Validation loss/accuracy of the generated models only after training when pressed
  • Open Test Folder : Used to select the path to the test dataset
  • Start Training : Starts the model training process when pressed
  • Augment Dataset : Augments the dataset when pressed
  • Save Trained Model : To save the model as model_name.pth file

Radio Buttons

  • Split required : Select this button IF and only IF there is no separate test dataset
  • Split NOT required : Default selection | Select this if you have both the training and test dataset, provide the path to both datasets by clicking Open folder and Open Test Folder buttons

Entry Text Boxes

  • Learning Rate : float | Set the Learning rate for training
  • Enter number of Channels : int | Select the number of channels in the given training image
  • Enter number of Classes : int | Enter the total number of classes to be classified in
  • Enter image shape : str | height x width | Should be a string and in the following format 32x32

Info

The GUI will run on the main thread and all the training and other reasonably heavy computation will be carried on different threads.

Danger

DO NOT CLOSE THE MAIN GUI WINDOW during the training process. This is not at all recommended, wait till the training process is over.

Results upon pressing Display graphs after completion of the training process

Alt text

Saving the model after training

Alt text

Select the path where you want to store the trained model

Alt text

To load the trained model

  • Select Load Model

Alt text

Attention

This demonstration was carried out in an old version of AutoNN hence the interface looks a bit different, but the functionalities are same so you needn't worry about that.