Computer Vision Onramp - Part 1 - Course Overview and Video Data

Computer Vision Onramp

Course Overview

Computer vision is about teaching computers to interpret and understand the contents of images and videos by automating perception. It’s about automating perception with cameras and other sensors. For example, identify bison in a scene or combining data from multiple cameras to create a digital reconstruction. Computer vision is where knowledge about the real world and digital data meet.

In this course, you will learn to use object detection to find, track, and count turtles in a video.

Workflow & Outline

Tracking Turtles with Object Detection follows a five-step workflow:

1. Import Video Data
2. Label Ground Truth Data
3. Train Detector
4. Detect Objects in Each Frame
5. Track Objects

Video Data in MATLAB

Create a Video Reader

Videos are sequences of images called frames, taken at fixed time intervals, with the frame rate stored as metadata. A video reader makes it easy to access both metadata and individual frames.

You can use the VideoReader function to create a video reader:

turtleVideo = VideoReader("turtles.avi");

The output holds information about the video file:

turtleVideo.Duration
turtleVideo.FrameRate
turtleVideo.Height
turtleVideo.Width
turtleVideo.NumFrames

Read Video Data

Read Frames Sequentially

Use the readFrame function to extract the next frame. The video reader tracks progress with the CurrentTime property. Each call advances the time by $1/framerate$.

frame = readFrame(turtleVideo);
imshow(frame)
t = turtleVideo.CurrentTime;

Read Specific Frames

The read function extracts a specific frame by its index:

frameMid = read(turtleVideo, 50); % Read 50th frame
imshow(frameMid);

You can skip to the last frame using Inf:

frameLast = read(turtleVideo, Inf);

To rewind, simply reset the CurrentTime property:

turtleVideo.CurrentTime = 0;