Weakly Supervised Point Cloud Segmentation via Conservative Propagation of Scene-level Labels

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Xia, Shaobo, Yue, Jun, Kania, Kacper, Fang, Leyuan, Tagliasacchi, Andrea, Yi, Kwang Moo, Sun, Weiwei
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915350265724928
author Xia, Shaobo
Yue, Jun
Kania, Kacper
Fang, Leyuan
Tagliasacchi, Andrea
Yi, Kwang Moo
Sun, Weiwei
author_facet Xia, Shaobo
Yue, Jun
Kania, Kacper
Fang, Leyuan
Tagliasacchi, Andrea
Yi, Kwang Moo
Sun, Weiwei
contents We propose a weakly supervised semantic segmentation method for point clouds that predicts "per-point" labels from just "whole-scene" annotations. The key challenge here is the discrepancy between the target of dense per-point semantic prediction and training losses derived from only scene-level labels. To address this, in addition to the typical weakly-supervised setup that supervises all points with the scene label, we propose to conservatively propagate the scene-level labels to points selectively. Specifically, we over-segment point cloud features via unsupervised clustering in the entire dataset and form primitives. We then associate scene-level labels with primitives through bipartite matching. Then, we allow labels to pass through this primitive-label relationship, while further encouraging features to form narrow clusters around the primitives. Importantly, through bipartite matching, this additional pathway through which labels flow, only propagates scene labels to the most relevant points, reducing the potential negative impact caused by the global approach that existing methods take. We evaluate our method on ScanNet and S3DIS datasets, outperforming the state of the art by a large margin.
format Preprint
id arxiv_https___arxiv_org_abs_2312_06799
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Weakly Supervised Point Cloud Segmentation via Conservative Propagation of Scene-level Labels
Xia, Shaobo
Yue, Jun
Kania, Kacper
Fang, Leyuan
Tagliasacchi, Andrea
Yi, Kwang Moo
Sun, Weiwei
Computer Vision and Pattern Recognition
Machine Learning
We propose a weakly supervised semantic segmentation method for point clouds that predicts "per-point" labels from just "whole-scene" annotations. The key challenge here is the discrepancy between the target of dense per-point semantic prediction and training losses derived from only scene-level labels. To address this, in addition to the typical weakly-supervised setup that supervises all points with the scene label, we propose to conservatively propagate the scene-level labels to points selectively. Specifically, we over-segment point cloud features via unsupervised clustering in the entire dataset and form primitives. We then associate scene-level labels with primitives through bipartite matching. Then, we allow labels to pass through this primitive-label relationship, while further encouraging features to form narrow clusters around the primitives. Importantly, through bipartite matching, this additional pathway through which labels flow, only propagates scene labels to the most relevant points, reducing the potential negative impact caused by the global approach that existing methods take. We evaluate our method on ScanNet and S3DIS datasets, outperforming the state of the art by a large margin.
title Weakly Supervised Point Cloud Segmentation via Conservative Propagation of Scene-level Labels
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2312.06799