SERVICESAI Data Processing Service

This is the most common method used in most industries especially retail, e-commerce, and healthcare by data scientists for data modeling. In this model 2d boxes are drawn close to the key objects in the image for classification, detection, and localization. Rectangular boxes are drawn with x and y-axis on the target object.

 Polygonal shapes are used to detect objects and location as most objects cannot fit into the rectangular box, by using complex and different polygon shapes better precession is achieved for data modeling of objects in different shapes and sizes.

This type of data modeling is used mostly on human body parts to recognize facial expressions, gestures, emotions by creating dots across the images and videos to detect even small variations between objects.

 A particular logic or pattern is used to classify objects within the image or video like in the below example pedestrians, trees, cameras, chairs, each object is segmented semantically for the machine learning algorithm to learn and is mostly used where an outdoor environment comes in for image/video annotations. Example: Self-drive cars to detect the objects in semantic patterns on the roads (outdoor environment).

These are similar to the bounding box model, but it captures additional information like mass, volume, and distance of objects, this is predominantly used for machine learning algorithms used in self-driving cars.

These annotations are used on roads with lanes for the self-driving cars to detect different lanes while driving on highways. It is simple as it eliminates unwanted noise and white space.