deep media organizer with deep leraning (mobilenet-ssd)

This is a nodejs package to organize files like images and videos in folders with respective classes detected by a Tensorflow object detect model (converted from python to js).

Usage

const DeepOrganizer = require('@nindoo/deep-organizer').DeepOrganizer

const modelConfig = {
    modelUrl: 'file://path/for/your/web_model/model.json',
    classes: {
        1: {
            name: 'CNH_F',
            id: 1,
            displayName: 'CNH_F'
        },
        2:{
            name: 'CNH_Fv',
            id: 2,
            displayName: 'CNH_Fv'
        }
    }
}

const mediaPath = 'media/path/videos-or-images'
const organizer = new DeepOrganizer(modelConfig, mediaPath)
organizer.loadModel().then(async ()=>{
    await organizer.organizeImagesTo(mediaPath)
    await organizer.organizeVideosTo(mediaPath)
})

modelConfig

const modelConfig = {
    modelUrl: 'It MUST start with file:// for local files or https:// for remote files',
    classes: 'It repesent your label_map.pbtxt from your tensorflow model'
}

Technical details for advanced users

This model is based on the TensorFlow object detection API. You can download the original models from here. We applied the following optimizations to improve the performance for browser execution:

1. Install the TensorFlow.js pip package:

pip install tensorflowjs

2. Run the converter script provided by the pip package:

The converter expects a TensorFlow SavedModel, TensorFlow Hub module, TensorFlow.js JSON format, Keras HDF5 model, or tf.keras SavedModel for input.

TensorFlow SavedModel example:

tensorflowjs_converter \
    --input_format=tf_saved_model \
    --output_format=tfjs_graph_model \
    --signature_name=serving_default \
    --saved_model_tags=serve \
    /mobilenet/saved_model \
    /mobilenet/web_model 

Tensorflow Hub module example:

tensorflowjs_converter \
    --input_format=tf_hub \
    'https://tfhub.dev/google/imagenet/mobilenet_v1_100_224/classification/1' \
    /mobilenet/web_model 

Keras HDF5 model example:

tensorflowjs_converter \
    --input_format=keras \
    /tmp/my_keras_model.h5 \
    /tmp/my_tfjs_model

tf.keras SavedModel example:

tensorflowjs_converter \
    --input_format=keras_saved_model \
    /tmp/my_tf_keras_saved_model/1542211770 \
    /tmp/my_tfjs_model

more information about convertion here