This GitHub is intended to gather all datasets, Python source code, and results of the paper above.
Semantic segmentation aims at assigning labels to every pixel of a given image. In the context of autonomous vehicles, semantic segmentation models should be trained with data collected from the traffic network through which vehicles are expected to circulate. Road regulation, weather conditions, and other context features may differ between regions, making local semantic segmentation datasets extremely valuable. However, the high ground truth annotation costs represent a hindrance to the development of such models. The upsurge of powerful feature learning architectures leaves room for semantic segmentation models trained on an unsupervised fashion. This observation vertebrates the purpose of this work: to produce coarse segmentation maps for scene understanding without the need of annotated data. We depart from an unsupervised model that yields low-quality results. The proposed methodology establishes a set of guidelines for the enhancement of segmentation maps. Obtained results expose an improvement of the segmentation quality thanks to the application of our devised guidelines, paving the way for the automatic generation of semantic segmentation datasets.
NOT AVAILABLE YET