Script for producing "neural network fooling images" with CPPNs. This is a simplified version of the algorithm described in Nguyen, Anh, Jason Yosinski, and Jeff Clune. "Deep neural networks are easily fooled: High confidence predictions for unrecognizable images." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,…

Script for Fooling Images as in Nguyen Yosinski Clune (2015)

/*
MIT License

Copyright (c) 2017 Kieran Browne

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

This script is designed to interact with Felix Andrews' CPPNX
https://floybix.github.io/cppnx/
It was built with browserify and run with Arturo Guzman's 'Custom 
JavaScript for websites' https://github.com/guzart/customjs
*/


const Clarifai = require('clarifai');
const $ = require('jquery');

const app = new Clarifai.App({
    apiKey: 'YOUR_API_KEY'
});

scores = [ // these also control which classes will be optimised for
    {name: 'train', value: 0},
    {name: 'cat', value: 0},
    {name: 'fish', value: 0},
    {name: 'house', value: 0},
    {name: 'starfish', value: 0},
    {name: 'baseball', value: 0},
    {name: 'remote control', value: 0},
    {name: 'guitar', value: 0},
    {name: 'tentacle', value: 0}]

function compare(a,b) {
    if (a.value < b.value)
        return 1;
    if (a.value > b.value)
        return -1;
    return 0;
}

writeScores = function() {
    if(scores.length > 9) scores = scores.slice(0,-1);
	  let scrs = scores.sort(compare).slice(0,8);
    $('#scores').remove()
    $('.cppnx-main-canvas').parent().prepend( 
        "<table id='scores'><tr>"+(scrs.map(s=>'<td><img style="margin-left: 2px; margin-bottom: 2px; width: 100px; height: 100px;" src="'+s.src+'"></td>').join(''))+"</tr><tr>"+(scrs.map(s=>'<td><p>'+s.value+'</p></td>').join(''))+"</tr><tr>"+(scrs.map(s=>'<td><p>'+s.name+'</p></td>').join(''))+"</tr></table>")
}

function options() {
    return $('img').not('#scores img');
}

let index = 0;

function predict() {
    let imgs = options();
    if(index>imgs.length-2) index = 0;
    else index++;
    console.log("trying ", index);

    return app.models.predict(
            Clarifai.GENERAL_MODEL,
            {"base64": $(imgs.get(index)).attr('src')
                .replace(/^data:image\/(.*);base64,/, '')},
            {selectConcepts: scores.map(x=>{return {name: x.name}})})
}

const updateScores = (res) => {
    return new Promise(function (resolve,reject) {
        let imageScores = res.rawData.outputs[0].data.concepts;
        let imgs = options();
        // regenerate options if no improvements exist
        let breed = false;

        for(let i=0; i<scores.length; i++) {
            let imageScore = $.grep(imageScores, x=>x.name == scores[i].name)[0]
            if(imageScore.value > scores[i].value) {
                console.log('new high score for '+scores[i].name);
                // add image src and new high score to scores array
                scores[i].value = imageScore.value
                scores[i].src = $(imgs.get(index)).attr('src')
                // update the scores table
                writeScores();
                // breed this image
                breed = true;
                // break; // don't continue checking scores
            }
        }
        if(breed) $(imgs.get(index)).click();
        else if(index == 0) $('.btn.btn-default.btn-lg').click();
    })}

main = () => {
    predict().then(updateScores)
    setTimeout(() => { main() }, 6000)
}

setTimeout(() => { main() }, 1000) // begin