Structural Segmentation of Visual Data

For understanding visual scenery, we construct deep-learning based AI algorithms for semantic segmentation of visual data. Segmentation here involves labeling of road scenes as “road”, “cars”, “sidewalk”, “pedestrian”, “bikes”, … which works by classifying all image pixels into the given labels/classes.
This figure consists of some example images from ITU Campus which are collected and processed by ITU Vision Lab. All of the examples are presented with their segmented versions.

Gozde Unal
ITU Vision Lab
Funded by
April 2018