171846d7af5ea91e63db508154eaffe8-Supplemental-Conference.pdf

Next we conduct more experiments on the generalization of unpaired, multi-category and multi-noise-ratio. Thefeatures of points in a point cloud are generated by the same MLP with the same weight. We add the random shiftxs [ smax,smax]3 to all the points of inputs. The last line is the evaluation results of our model trained on the dataset that contains singlenoiseratio. Unsupervised point cloud object cosegmentation by co-contrastive learning and mutual attention sampling.