Tfjs Evolution

This is a placeholder for a package I am planning to publish. I have experimented with Evolutionary Computing in TensorFlow.js, and I am planning a paper and a NPN package. I was unable to find a NPM package, and decided to create one!

I am releasing the parts by parts. You can start to test the released parts, and let me know your thoughts.

Seeing your images

You can see your images, it can useful to have an idea about how it looks like, what you are trying to feed the model. Remember: "garbage in garbage out".

Import the directive for the panel.

import { DisplayPanelComponent } from 'tfjs-evolution';

Make sure to import the component. Angular is now standalone by default.

imports: [DisplayPanelComponent]

Create a local instance, for calling the component method.

@ViewChild(DisplayPanelComponent) child!:DisplayPanelComponent;

Add a list of class names. It must be the same on the folder names.

  classes_names=[
    "Canário-da-Terra",
    "Pardal",
    "Bem-Te-Vi",
    "Suiriri-Cavaleiro",
    "Lavadeira-Mascarada",
    "Anu-Branco",  
    "Beija-Flor-Tesoura", 
    "Casaca-de-Couro-da-Lama",
    "Garibaldi",
    "Periquitão",
    "Pombo-Doméstico",
    "Quero-Quero",
    "Rolinha-Roxa",
    "Sabiá-Do-Campo",
    "Urubu-de-Cabeça-Preta"
  ]

Make the call for loading images. You could most likely make the call at the constructor.

ngAfterViewInit(): void {
    const options={ 
      base: "./assets/dataset",//base where are your iamges
      file_name:"image",//how your images are named
      file_extension:"jpeg" //extension used
    };

    //Finally, call the load images methods
    this.child.loadImages(5, options)
  }

Finally, do not forget to add the HTML code

<h1>Images used on the training</h1>
<neuroevolution-display-panel></neuroevolution-display-panel>

You should see in our HTML file, in Angular, the images

Updates

I have finished! Hope to publish a paper soon!

GitHub repo

You can find working examples on: https://github.com/JorgeGuerraPires/transfer-learning-by-evolutionary-computing/tree/master

Just clone the repository, NPM install and have fun!

Further help

Feel free to get in touch: [email protected]