Performance Prediction
Published:
Developing perception systems for autonomous driving has not been a trivial task. Besides of having been one of the central topics in the community, there are just so many scenarios in the streets to consider in order to secure the system’s safety. And while sensor devices are getting better and cheaper, a system that robust against noise or corruption needs to be studied further. To locate objects and boundaries (lines, curves, etc.) in street or traffic scenes a typical approach is to build a system that transforms the input into something more practical to analyse. Object detection locates existing objects in a scene and assigns a category to each identified object, while semantic segmentation deals with a pixel-level classification, producing a mask of certain classes.