One of the reasons why people love creating a text journal is because text can be searched. This means that, even if your journal is a thousand page long, with just a few keyboard strokes, you can see where in the document what you are looking for is located.
With Google Cloud AI, searching through scenes from a video recording is now possible. This means that, you no longer need to spend hours looking through an hour-long video, just because you want to re-watch a 1-minute scene.
But is building such an AI that analyzes and categorize scenes from a video really possible? Yes.
This AI Already Exist in Google Photos
If you use Google Photos, you would have noticed that when you use the search box to search for a picture, the result it will return is not only pictures that have the search term in their file name. it will also return pictures that the built-in AI of Google Photos thinks it is in the picture.
For example, if you searched for the term Wedding. The result that will be shown will not only pictures have Wedding in it file name, but also pictures that the Google Photos AI thinks it is a wedding picture.
Isn’t that just amazing? Now, however, building an AI that does a similar thing with videos is possible.
Building an AI that Analyzes a Video is Possible
I recently discover a post by Dale Markowitz, an Applied AI Engineer with Dale on AI, where he explained that he was able to build an AI that analyzed and categorized his 30 years old family videos which is more than 456 hours of footage.
With the AI she built, she was able to see footage that captured her taking her first steps as a child, and also a footage of when she was singing “It’s Not Easy Being Green” in her kindergarten’s production of the Sesame Street musical.
How the AI was Built.
Dale Markowitz explains how she was able to build the AI that analyzed the family videos, in an in-depth technical blog post where she not only explain the process but also shows you all the code and guide you through the process.
Here is a video she made explaining how she made the AI as a father’s day gift for her dad and his response. Enjoy.
The Process in a Glance
Dale explains that what she used to build this AI are, Google Cloud, Google Video AI, Algolia, and Flutter to build the project.
Google Cloud is for storing the videos in the cloud, since the project happened in the cloud.
Google Video AI is for analyzing the video so as to create contents that can be searched. Example, creating audio transcript of the video, extracting time stamp and other onscreen text. You can also use it to split the videos scenes.
Algolia is for making those transcript, time stamps and onscreen texts searchable.
Flutter is used to build the front-end used to access the database.
Dale, says that the guide she has put up will enable you to build your own AI even if you an AI genius. However, if you want to build your own AI, but you are having difficulty doing so, Dale encourage you get touch with her via Twitter.
She is very excited about this project of hers, so I think she will be more than willing to answer your reasonable questions.
Do you find this article helpful? Your Friend might too. So, please Share it with them using the Share button above.
Will you like to get notified when I post new updates? Then Follow me on any of my social media handles: Google News, Telegram, WhatsApp, Twitter, Facebook, Pinterest.
You can also drop your email address below if you wish to be notified by mail.