Fig. 3From: Pseudo-labelling-aided semantic segmentation on sparsely annotated 3D point cloudsGradual semantic labelling of the scene with kdist. The top row shows accepted pseudo-labels after a 8, b 80, and c 105 iterations. The unweighted average F-scores at the iterations are a 95.0%, b 91.1%, and c 74.2%. The rightmost image shows the final semantic labelling. Colours correspond to semantic classes. White dots indicate initially annotated points. The bottom row highlights errors in red while correct labels are shown in white. In both the top and the bottom images, unlabelled points are shown in greyBack to article page