0:00
Cloud Video Intelligence is an API that lets you understand your videos entities at shot,
frame, or video level.
Let's see what you can do with the Video Intelligence API.
First, the video API lets you do label detection,
which tells you what's in this video and it does this in two ways.
At a high level, it'll tell you what is your video about,
and in a more granular level,
it can tell you what's happening in every scene of your video.
The API also lets you do Shot Change Detection.
So, for example, if your video changes from
a landscape pan to a close up of a person doing an interview,
it will give you the time stamps every time the camera changes shots.
It also lets you do Explicit Content Detection.
This can identify inappropriate scenes in your video.
And finally, the video API supports Regionalization,
which lets you specify the region where your video API requests should be executed.
The best way to see how the video API works is through a live demo.
Let's take a look. Here, we have
a Superbowl commercial for Google Home and
I'm just going to play the first couple seconds of the video.
We can see here at the video has a lot of scene changes.
It started with the mountain pass then it moved to a city street.
Now we see a dog and then a garage.
You get the idea.
If we were to have to manually classify this video,
we need to watch the entire thing,
write down what was happening in each scene,
and then store these tags in a database somewhere.
With the video API, we can do this with a single REST API call.
If we look below the video,
we can see a visualization of the JSON response we get back from the video API.
It knows there's a dog in the video and it knows
exactly which seconds the dog can be found.
It also knows there's a birthday cake at the end of the video.
If we scroll down,
we can see some more of the labels that are returned.
We can see it's able to identify the exact breed of dog from that opening scene,
and it's even able to identify the mountain pass from the beginning.
This is what the video API can do with one video,
but you likely have more than one video that you want to analyze.
Let's say, you're sports media company and you've hours of
sports footage but you want to create a highlight reel of only your baseball videos.
Doing this manually would take hours.
With the video API, this is easy using the JSON response you get back.
I have a library of videos here with a number of
sports videos and I want to find only my baseball clips.
I'm going to search my library for baseball, and boom.
We see that the video API is immediately able to
identify which of our videos contain baseball.
It's also able to identify where in these videos we can find baseball clips.
This is my favorite example,
this is the year end search video that Google publishes
at the end of every year highlighting top searches.
It's obviously got a lot of different types of clips in it.
It only has one or two seconds that contain a baseball clip.
With the video API, we're able to jump right to that point in the video.
This is from 2016,
when the Cubs won the World Series.
So let's do one more search.
Let's find all of the Android clips in our video.
Here, we can jump to all the points in our videos where an Android phone is present.
We can see that what used to take hours,
now takes minutes or seconds with the Video Intelligence API.
Google Cloud Training
