All News | Boards | Chips | Devices | Software | Archive | About | Contact | Subscribe
Please whitelist in your ad blocker. Without ads from our sponsors, we cannot continue publishing this site. Thanks :-)

Google releases Cloud Vision API with demo for Pi-based robot

Dec 4, 2015 — by Eric Brown — 3,642 views

Google released a preview of its Cloud Vision API for tasks like identifying objects and faces, plus a Linux demo that runs on a Raspberry Pi-based robot.

Some of the image analysis wizardry used by Google services, such as Google Photos, is now available to developers. Google is offering a free limited preview of its Google Cloud Vision API, which is available as an “easy to use” REST API, says the company. Google also released demo code using the API that turns the Raspberry Pi-based Dexter Industries GoPiGo robot or any other camera-enabled robot based on the Pi into an image recognition and analysis bot.

Annotations returned by Cloud Vision API on images
(click images to enlarge)

The Google Cloud Vision API, which is based on the Google-backed open source TensorFlow machine intelligence library, integrates machine learning models that can quickly classify images into thousands of categories. The API can also detect objects and faces within images, although Google notes that it does not enable facial recognition, and promises that no facial IDs are stored on Google servers. Other applications include building metadata from an image catalog, identifying offensive content, or analyzing “image sentiment” for marketing purposes, says Google.

Cloud Vision API features are said to include:

  • Label/Entity Detection — Identify the dominant entity within an image, and build metadata based on an image catalog for image based searches or recommendations.
  • Optical Character Recognition — Retrieve and reproduce text from an image with automatic language identification supporting multiple languages.
  • Safe Search Detection — Detect inappropriate content within an image (based on Google SafeSearch).
  • Facial Detection — Detect faces and facial features such as eye, nose and mouth placement, plus identify the likelihood of over eight attributes like joy and sorrow.
  • Landmark Detection — Identify popular natural and manmade structures, along with the associated latitude and longitude.
  • Logo Detection — Identify product logos.

The Cloud Vision API will evolve over time as new concepts are introduced. For example, Google plans to support integration with Google Cloud Storage. One early beta tester was Aerosense, a professional fixed-wing drone developed by Sony and ZMP.

GoPiGo (left) and testing the Cloud Vision API’s emotion recognition capabilities
(click images to enlarge)

The Raspberry Pi robot demo application consists of a “few hundred” lines of Python code that calls the Vision API, says Google. The YouTube video below shows the wheeled GoPiGo bot running the demo, but it’s said to work with any Raspberry Pi based robot. The demo enables a robot to identify objects and basic emotions as it rolls about. You can even teach the robot to follow a face.

GoPiGo running Cloud Vision API

Further information

The limited preview of Google’s Cloud Vision API, as well as the Raspberry Pi robot demo, are available for free download. More information may be found in the Google Cloud Blog announcement of the release, as well as Google’s Cloud Vision API product page.

(advertise here)


4 responses to “Google releases Cloud Vision API with demo for Pi-based robot”

  1. santa2015 says:

    where can I find the Raspberry Pi robot demo? I couldn’t find it on their website (

  2. Todd S says:

    I waited a bit for the demo code, couldn’t find it anywhere but there are other examples for google cloud vision api (on github). At Google I/O this past week they demoed this again, so maybe there will be more pressure to put the code up online.

Please comment here...