Titanium Deep-Learning

Use the JetPac DeepBeliefSDK framework in Appcelerator Titanium.

Note: This module does not support apps running with app-thinning enabled so far. It will be part of upcoming versions and is scheduled already.

Example

var TiDeepLearning = require('ti.deeplearning');

TiDeepLearning.initializeNetwork({
    name: 'jetpac.ntwk' // Search and download from the DeepBelief SDK Github page
});

TiDeepLearning.classifyImage({
    image: 'macintosh.jpg',
    minimumThreshold: 0.01,
    decay: 0.75, 
    callback: function(e) {
        Ti.API.info(e.result);
    }
});

The above example with return an array with predictions, like this:

(
    {
        label = printer;
        value = "0.1368110626935959";
    },
    {
        label = monitor;
        value = "0.06563640385866165";
    },
    {
        label = "desktop computer";
        value = "0.143329456448555";
    },
    {
        label = screen;
        value = "0.4579531848430634";
    },
    {
        label = iPod;
        value = "0.01504476089030504";
    },
    {
        label = "cash machine";
        value = "0.02499096281826496";
    },
    {
        label = safe;
        value = "0.01404030714184046";
    },
    {
        label = "entertainment center";
        value = "0.01761411875486374";
    },
    {
        label = television;
        value = "0.06137070804834366";
    }
)

Note that the values highly depend on your network and complexity of classification. For a full example, check the demos in example/app.js.

Author

Hans Knoechel (@hansemannnn / Web)

License

MIT

Contributing

Code contributions are greatly appreciated, please submit a new pull request!