Running Tensorflow on RICOH THETA V

I’ll show you how I successfully installed Tensorflow as a RICOH THETA V plug-in

Application Overview

I’ll show you how to use four Tensorflow demo applications running on the RICOH THETA V.

  1. Automatically classify objects in the THETA range of view
  2. Detect objects such as people
  3. Stylize images
  4. Talk to the THETA and activate demo buttons on the screen

Requirements

  • RICOH THETA V in developer mode (it’s free to put the camera in developer mode)
  • theta-tensorflow-demo.apk (or compile it yourself from GitHub)
  • Vysor (used to set app permissions)
  • adb, part of Android Studio

I’ll cover the basic installation at the end of this article. For more information on working with plug-ins, see the Plug-in Developer Guide.

Demo Overview

The demos below are shown running directly on the RICOH THETA V. The screenshots are from Vysor, which is a tool to show a virtual Android screen from the THETA V on your computer. The demos are based on the workshop created by Makoto Shohara. He modified the demos from the Tensorflow Android demos.

Tensor Flow Detect

correctly identify person. It thinks the computer monitor is a TV.

The application automatically takes a picture.

Testing Tensorflow Stylize

Testing Tensorflow Speech

I say, “Left” to the THETA V and Tensorflow detects it.

Tensorflow Classify

Installation

  1. Download theta-tensorflow-demo.apk or compile from GitHub repo
  2. Connect your THETA V to your computer with a USB cable
  3. Run the command adb install theta-tensorflow-demo.apk

Set Plug-in Permissions

  1. Start Vysor
  2. Go into Settings
  3. Go to Apps
  4. Select TensorFlow Demo
  5. Select App permissions
  6. Enable permissions as shown in the screenshot below

Run Applications

There are four applications you can use. With the camera still connected to your computer with a USB cable, click on the application you want to run in Vysor.

The camera will automatically start to run. The lens opposite the shutter button is the center of the equirectangular image.

Additional Information

Sign up for the Nov 10 RICOH THETA Microconference now