5. MetaCloud

MetaCloud is the MbientLab cloud platform used for raw sensor:

  • Data Storage
  • Data Visualization
  • Data Analysis
  • Pattern Recognition (Machine Learning)

MetaCloud is available as a monthly subcription and works only with MetaSensors.

In this tutorial we will look at how to create an account, upload sensor data, and use the cloud machine learning tools.

5.1. Subscription

To create an account, follow the steps:

  1. Navigate to the cloud login page: https://metacloud.mbientlab.com/
_images/metacloud-8.png
  1. Select “Start a Free Trial”
_images/metacloud-9.png
  1. Create a username and password
  2. Verify your email

Once your account has been created, verify your email address. Go to your inbox and make sure to acknowledge your new MetaCloud account. If you didn’t receive an email, check your spam/trash folder. Once your email is verified, simply login again.

  1. Log in

You will automatically be directed to the “Profile” page in order to select a MetaCloud subscription tier.

_images/metacloud-10.png
  1. Select a subscription and initial a payment
_images/metacloud-11.png

Select your desired payment plan and input your credit card information. All subscriptions start with a 3 day FREE trial. You will not be charged for the first 3 days of your MetaCloud subscription.

Please note you will be billed monthly and you can cancel at any time.

If everything works, you will receive a notification of a successful payment. You can check your current subscription status on the subscription page at anytime.

5.2. Cancellation

You can cancel your subscription at any time by login in to your account and navigating to the “Profile” page.

In the “Settings” page of the “Profile” tab, you will find a “Cancel” Button.

You will get a pop-up confirming your cancellation.

5.3. Data Storage

To upload data to MetaCloud:

  1. Download and use the MetaBase App
  2. Acquire sensor data with MetaBase
  3. Upload the data to the MetaCloud with MetaBase (it will prompt you to login)
  4. Verify that your data is on the cloud

First, let’s look at how to update data in this video:

You can view your raw data on the “Explorer” page on the “Home” tab.

_images/metacloud-4.png

There is a delete and download button for each data set and session. The data is available in CSV format.

Please note that your devices and data are automatically managed by the cloud; you don’t need to change, upload or update any of the sensor data or information manually.

5.4. Data Visualization

There are two sections on your MetaCloud Dashboard, your Main page “Home” and the “Data Analytics” section.

The “Dashboard” page on your “Home” tab is where all your data can be visualized.

There is a data picker for which you can select data by date, device mac, or session name:

_images/metacloud-6.png _images/metacloud-7.png

If there are multiple sensors for that data session, each type of sensor data will show up in a separate graph.

_images/metacloud-12.png

You can select any portion of the graphs to zoom in on that section. This is useful for exploring areas where data is dense.

_images/metacloud-13.png

Press the reset button at the top of each graph to restore the graph to its original zoom factor (highlighted in red below).

Press the download button on top of each graph to download the data for that sensor in the data session (highlighted in blue below).

_images/metacloud-14.png

5.5. Data Analysis

The “Data Filtering” page on your “Data Analytics” tab is where all your data can be filtered and analyzed.

Once data has been selected, it will be graphed on your screen automatically. By default the latest data set is always graphed on the main page.

You are free to zoom and refresh the graphs. Select any filter you want to apply to the data. Currently, we allow stacking of up to 2 filters at the same time. This is very useful for doing a very quick visual inspection of the data.

In the example below, we combine 2 filters to quickly analyze walking data using the accelerometer:

  • First we apply a “low pass” filter. This removes a lot of the random noise that is common in raw accelerometer data.
  • Then we apply a “root mean square” filter. This lets us look at the magnitude (or total acceleration in all directions), which gives us a very crude estimate of how fast we were moving.
_images/metacloud-5.png

From the filtered graph, we can quickly identify the areas where there was significant movement and how large/fast those movements were.

5.6. Pattern Recognition (Machine Learning)

The “Machine Learning” functionality in MetaCloud allows you to train the computer to identify certain patterns. This is a 2 step process that involves:

  1. Feeding the machine a sample data set with “examples” of the pattern that you want to identify.
  2. Feeding the machine test data sets, where the machine will use the “examples” learned from step 1 to identify similar patterns.
_images/metacloud-3.png

To perform this operation:

  • First, select a sample training data set; a graph will appear. Use your cusor to “highlight” the areas which you want the machine to learn as a “pattern”.
  • Click “train”.
  • Once trained, you will be asked to select a test data set. Ideally, this should be a data set that contains similar patterns, but is NOT the same set as the training data.
  • Click “test”. The machine should automatically identify all patterns which are similar to the ones you selected in the training set.