Summary

This was our first advanced computer vision topic where we used statistical and deep learning-based object tracking models to solve case studies involving object surveillance and other similar examples. We started with a very basic object tracker that was based on a very naive approach. We also saw how the tracker, while computationally very cheap, has very low performance and can't be used for most real-life scenarios. We used the HSV color space in Exercise 6.01, Object Tracking Using Basic Image Processing, to track a lemon across various frames on the input video. We then went ahead and discussed common non-deep learning-based trackers – the Kalman filter and the meanshift and CAMshift filters. Next, we discussed the OpenCV object tracking API, where we listed the eight commonly used object trackers and their pros and cons. We also had a look at how we can use those trackers and how GOTURN requires a slightly different process since we need to download the Caffe...