We open up "cloud machine learning" used by our company in general, it is easy to make such a big deal

GoogleGCP NEXT 2016So, Google's cloud-based machine learning platform "Cloud Machine Learning"publish. We have made it possible for application developers to use powerful machine learning functions with their own services.

Google Cloud Platform Blog: Google takes Cloud Machine Learning service mainstream
https://cloudplatform.googleblog.com/2016/03/Google-takes-Cloud-Machine-Learning-service-mainstream.html

Google Cloud Machine Learning at Scale - Google Cloud Platform
https://cloud.google.com/products/machine-learning/

Google uses cloud machine learning to do Machine Learning (machine learning) on ​​the cloud for its own service development. For example, Google voice search, Google Now suggestion feature, Google photo image classification function, Google translation, email applicationInbox reply auto generation functionEtc. are being developed by cloud machine learning. Cloud machine learning function supporting this Google service is now available to general users as "Cloud Machine Learning".

In Cloud Machine Learning, Google developed machine learning model is provided as an API, which enables developers to use the following powerful machine learning models immediately for service.

· Image Analysis API "Cloud Vision API"
Cloud Vision API is an API that can analyze images captured by camera with Google's cloud service "Google Cloud Platform".

· API for voice recognition "Cloud Speech API"
Also released "Cloud Speech API" which can use speech recognition function on Google Cloud Platform.

The following movies explain what you can do with Cloud Vision API and Cloud Speech API.

What is Cloud Speech API? - YouTube


With the Cloud Speech API, developers can use the latest speech recognition technology developed by Google as an API. For example, voice data sent from an application is converted back to text in the cloud and returned.


The Cloud Speech API uses common machine learning techniques in common with voice search and voice input of Android terminals.


When I input voice "What is the weather in Tokyo?" ...


It is also possible to return the current temperature of Tokyo.


The Cloud Speech API already supports 80 different languages ​​and has a high cognitive ability that is not affected by a little noise. It's easy for application developers to use advanced voice recognition models and interactive features developed by Google for many years.


So let's demonstrate what you can do with the Cloud Speech API with this robot "GoPiGo".


GoPiGo is a Raspberry Pi-based robot, which carries a camera and microphone and can recognize voice and images.


Audio data acquired by GoPiGo with a microphone ......


Sent to the cloud, you can receive recognition results in real time as text.


Since conversion to text data can be processed in parallel with speech recognition, cloud can be processed in real time even if it is sent continuously without interrupting speech.


"Can you dance?"


GoPiGo recognized the voice "Can you dance?" And started dancing as the answer.


The API can also specify the language to recognize.


When speaking to "turning right" in Japanese, I could recognize the words without problems.


"Portable left" in Portuguese is also OK.


The API can also be combined with other APIs in Google Cloud. For example, let's combine the Cloud Vision API of the image recognition API with the OCR function.


If you show GoPiGo the sentence "left look, right look, turn around" ....


Recognize the image with Cloud Vision API and convert it to text ......


It started to move accordingly.


The Cloud Speech API enables batch processing as well as real-time stream recognition.


It can also be used for transcription of voice recorded by voicemail, podcast, call center, etc.


It is also possible to create an interactive voice recognition application that leverages the powerful machine power of Google Cloud Platform.


· Large machine learning function
Cloud Machine Learning also has a big data analysis function that finds purpose from a lot of big data. Using this, it is possible for the developer to design and manage a large-scale machine learning model that analyzes big data. Of course, because we use Google Cloud Platform with powerful machine power, data analysis is short and accurate. You can also integrate with other cloud data platforms like Google Cloud Storage and Google's BigQuery.


With the addition of Cloud Machine Learning to Google Cloud Platform that makes it easy for application developers to leverage the machine learning development environment of Google, it seems to be expected that new applications completely different from the past will appear one after another.

Google Cloud Platform including Cloud Machine Learning can be tried for free by registering from the following page. For the 60 days of the free trial period, you can use the function of 300 dollars (about 34,000 yen).

Free trial of Google Cloud Platform
https://console.cloud.google.com/freetrial