With rapid strides being made in the field of image search and classification as well as Computer Vision Algorithms for Images, we are but years away from making an exciting breakthrough that will open the gates to human-like AI. There are many interesting problems inherent in the field of computer vision. This can range from simple Image classification to image data extraction and search analytics.

Classification

This is possibly one of the most studied problems in computer vision. This consists of separating an image into one of many possible categories and this is done with the help of datasets like ImageNet, which has millions of classified images. In the last few years, these classification models have surpassed human performances, and the problem is considered to be practically solved.

Localization

This involves finding the exact location of a single object in a picture frame and is comparable to image classification. Localization is helpful in lots of real life problems, and one of its applications is in ‘smart cropping’. Localization can also be integrated with image classification for location of the object and placing it into one of many categories.

Instance Segmentation

Instance segmentation is when we would want to not only detecting objects within images but also to find a pixel per pixel mask for each of the images which are detected. This problem is known as object segmentation or ‘instance’.

Object Detection

When we have localization as well as classification at the same time, it leads us to the great need for detecting multiple objects at the same time. Object detection deals with finding as well as classifying a variable number of objects in an image. The important bit to remember is the “variable”. In stark contrast to classification, the o/p of object detection will be variable as there can be any number of objects in an image.

These techniques are poised to usher us into a new age where Deep Learning is the norm rather than the exception, exciting times lie ahead. Though this is a technical thing to understand but something one should try to get knowledge on.