Practical AI on the Google Cloud Platform

Chapter 6. Video Analysis on the Cloud

With the improved developments in image analysis, it was only a matter of time before we would be using the same techniques to analyze video. If we ignore the audio, video is for the most part just a stack of pictures—pictures that are in a given sequence that describes some order of events or context. As we learned from NLP, context can matter, and in video it certainly matters.

In this chapter we look at the process of analyzing video for a variety of applications, ranging from video indexing for capturing or tagging content in videos, to using the event sequence itself to capture the activity. Video indexing, while just an extension of image analysis, has wide-ranging use cases, from security to streaming. Indexing video is like asking the question “Who are they?” while identifying motion or action in video can answer the question “What are they doing?”

In this chapter we will first look at how to load and analyze video with Python on Colab. Then, we look at applications of AI with respect to video, in particular the task of automatic video indexing. From video indexing, we will move on to using a webcam to detect faces. Finally, we finish on a practical example in which we use a TF Hub human-motion detector to identify human activity in videos.

The following is a list of the main topics we will cover in this chapter:

• Downloading Video with Python

• Video AI and Video Indexing

• Building a Webcam Face Detector

• Recognizing Actions with ...