ViewPCL: a point cloud based active learning method for multi-view segmentation

TL;DR

ViewPCL introduces a novel active learning framework for multi-view semantic segmentation that uses a discrepancy-based score on point cloud distributions, leading to more data-efficient and explainable segmentation models.

Contribution

It presents a new active learning method leveraging geometrical discrepancies in multi-view point clouds, enhancing data efficiency and interpretability.

Findings

Improved segmentation performance with less labeled data

Effective discrepancy measure for view-based active learning

Open-source implementation available

Abstract

We propose a novel active learning framework for multi-view semantic segmentation. This framework relies on a new score that measures the discrepancy between point cloud distributions generated from the extra geometrical information derived from the model's prediction across different views. Our approach results in a data efficient and explainable active learning method. The source code is available at https://github.com/chilai235/viewpclAL.