Classifying and segmenting similar object types with computer vision
Using the techniques of semantic segmentation and instance segmentation, you can identify and label all the pixels in a video frame that relate to a particular type of object (semantic segmentation). The next step is to be able to recognize, uniquely identify and label different instances of the same type of object (instance segmentation).
For example, semantic segmentation gives us the ability to determine that there are dogs in an image and segment that object from others. Instance segmentation then allows us to determine that there are multiple types of dogs within a frame, segmenting and categorizing them as sub-variants of the same object type.When it comes to video analytics, popular uses of object classification include:
Manufacturing
Automating tasks on the factory floor and monitoring output on production lines with computer vision requires the ability to detect and segment objects as well as categorize multiple objects of the same type.
eCommerce & virtual retail
Enables shoppers to test different types of make-up, try-on clothes, and even upload pictures of outfits and automatically trigger a catalog search for similar garments.
Self-driving cars
Use pixel-perfect labeling to detect and distinguish objects from each other – such as identifying lanes, pavement, other vehicles and pedestrians.
*Disclaimer: Outcomes described and depicted are illustrative in nature. Learn more.