In this chapter, two different computer vision problems were shown. In segmentation, both the pixel level as well as convolutional neural net-based methods were shown. FCN shows the effectiveness of segmenting an image using the feature extraction method and, as a result, several current applications can be based on it. In track, two different approaches were discussed. Tracking by detection and tracking by matching can both be used for applications to track objects in the video. MOSSE tracker is a simple tracker for fast-paced applications and can be implemented on small computing devices. The Deep SORT method explained in this chapter can be used for multi-object tracking that uses deep CNN object detectors.
In the next chapter, we will begin with another branch of computer vision that focuses on understanding geometry of the scene explicitly. We will see methods to...