Join the Bot Libre open source project at BotLibre.org
Docs : User Manual : Deep Learning & Analytics

Deep Learning & Analytics

Overview

An analytic is a neural network, deep learning model, or analytical model that can be used to predict or classfiy data such as images, audio, text, and more.

The analytics library provides many open source examples of analytics for image classifcation and other usages. You can create your own analytic and upload a binary network file such as a python .pb file that represents your model.


Creating

You can create your own analytic from the browse analytics page. To create a analytic, just give it a name, description, and category to categorize it under. You can either make the analytic private and choose who can access it, or make it public and accessible by anyone.

Properties

Properties Description
Analytic Name Enter a analytic name.
Description Optional description.
Details You can enter optional additional information.
Disclaimer You can enter optional legal information.
License Optional license to release the analytic and all of its content under.
Website If this analytic has its own website, you can enter it here.
Website Subdomain (or domain) You can choose a subdomain to host your analytic's own website, or give a domain that you have registered and forward to this server.
Categories Comma separated list of categories to categorize the analytic under.
Tags Optional comma separated list of tags to tag the analytic under.
Private A private analytic is not visible to the public, only to the user and users grant access.
Hidden A hidden analytic is not displayed in the browse directory.
Access Mode Define who can access this analytic.
Fork Access Mode Define who can fork(copy) this analytic.
Content Rating Rate the analytic.
Ad Code You can display ads on your analytic's pages.

Analytic Network

The analytic network editor allows you to configure and to upload or import the required files for both graph and labels for your analytic. Currently the analytic network binary file must be a python Tensorflow .pb file. Support for other network types will be added in the future. The labels file is a text file that matches the binary network's output.




Network

Properties Description
Analytic Type Types of analytic (inception_v3, mobilenet_0.25, mobilenet_0.50, mobilenet_0.75, mobilenet_1.0)
Analytic Image Size Set the image size of the trained images
Analytic Feed The input name used while training the images
Analytic Fetch The output name used while training the images

Network (Buttons)

Icon Property Name Description
Save Save the current changes.
Upload Upload graph .pb file.
Download Download the current uploaded graph.
Delete Delete the uploaded graph.

Labels

Properties Description
Text Area Displaying the labels saved in the uploaded text file.

Labels (Buttons)

Icon Property Name Description
Save Save the current changes.
Upload Upload graph .pb file.
Download Download the current uploaded graph.


Custom Tensorflow and Bot Libre Analytic Networks

The custom analytic network editor allows you to configure and to upload or import the required files for both graph and labels for a custom analytic. Currently the analytic network binary file must be a .h5 file for a custom Tensorflow network, or a .ser file for a Bot Libre network. There should be no labels or properties. Custom networks can be trained and tested using the Web API.




Network

Properties Description
Input Nodes Number of input nodes
Output Nodes Number of output nodes
Intermediate Nodes Number of nodes in each intermediate layer
Intermediate Layers Number of intermediate layers
Activation Function The activation function used for each layer

Network (Buttons)

Icon Property Name Description
Save Save the current changes.
Reset Network Reset the custom network by making a new network with the configurations above.
Upload Upload graph .h5 or .ser file.
Download Download the current uploaded graph.
Delete Delete the uploaded graph.


Analytic Repository

The Analytic Repository allows you to add a label and import images inside one of each label labels.




Labels

Icon Property Name Description
Delete Delete the selected label.
Add Add a label

Properties Description
Labels Select a label



Add Label

Properties Description
Add label Add a label to the list

Analytic Training

Training images.




Training Network

Properties Description
Start Training will start based on availability. This might take several hours.

Analytic Testing

Testing images.




Test Network

Properties Description
Start Testing will start based on availability.

See Also