Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling

arXiv:2309.12378v2 Announce Type: replace
Abstract: Traditionally, training neural networks to perform semantic segmentation required expensive human-made annotations. But more recently, advances in the field of unsupervised learning have made significant progress on this issue and towards closing the gap to supervised algorithms. To achieve this, semantic knowledge is distilled by learning to correlate randomly sampled features from images across an entire dataset. In this work, we build upon these advances by incorporating information about the structure of the scene into …

abstract advances algorithms annotations arxiv correlation cs.cv feature gap human issue knowledge networks neural networks progress sampling segmentation semantic through training type unsupervised unsupervised learning