Weakly Supervised 3D

Weakly supervised 3D scene understanding aims to train accurate 3D segmentation models using minimal labeled data, addressing the significant cost and time associated with full annotation. Current research focuses on leveraging techniques like contrastive learning, probabilistic methods for pseudo-label generation, and incorporating information from multiple modalities (e.g., RGB images and text descriptions) to improve segmentation accuracy from sparse annotations. These advancements are crucial for expanding the applicability of 3D scene understanding to diverse domains, including medical imaging and robotics, where fully labeled datasets are often unavailable. The resulting data-efficient methods promise to significantly reduce the annotation burden and accelerate progress in various fields.

Papers

July 13, 2024

3D Weakly Supervised Semantic Segmentation with 2D Vision-Language Guidance
Xiaoxu Xu, Yitian Yuan, Jinlong Li, Qiudan Zhang, Zequn Jie, Lin Ma, Hao Tang, Nicu Sebe, Xu Wang
Semantic Segmentation 2 Dimensional 3D Semantic Segmentation 2 Dimensional Vision Language Model Weakly Supervised 3D Characteristic Guidance

April 19, 2024

Contrastive Gaussian Clustering: Weakly Supervised 3D Scene Segmentation
Myrna C. Silva, Mahtab Dahaghin, Matteo Toso, Alessio Del Bue
Segmentation Mask 3D Segmentation Scene Segmentation Contrastive Clustering Weakly Supervised 3D

March 5, 2024

Enhancing Weakly Supervised 3D Medical Image Segmentation through Probabilistic-aware Learning
Zhaoxin Fan, Runmin Jiang, Junhao Wu, Xin Huang, Tianyang Wang, Heng Huang, Min Xu
Medical Image Segmentation Pseudo Label Uncertainty Aware Multi Head Self Attention Segmentation Loss 3D Medical Imaging Weakly Supervised 3D

December 27, 2023

Multi-modality Affinity Inference for Weakly Supervised 3D Semantic Segmentation
Xiawei Li, Qingyuan Xu, Jing Zhang, Tianyi Zhang, Qian Yu, Lu Sheng, Dong Xu
Point Cloud Segmentation Scene Level Label Weakly Supervised 3D Affinity Learning

December 3, 2023

A Review and A Robust Framework of Data-Efficient 3D Scene Parsing with Traditional/Learned 3D Descriptors
Kangcheng Liu
Point Cloud 3D Scene Robust Framework Point Cloud Understanding 3D Object Detection Benchmark Weakly Supervised 3D 3D Descriptor

May 25, 2023

All Points Matter: Entropy-Regularized Distribution Alignment for Weakly-supervised 3D Segmentation
Liyao Tang, Zhe Chen, Shanshan Zhao, Chaoyue Wang, Dacheng Tao
Pseudo Label Exceptional Point Distribution Alignment Weakly Supervised 3D

March 26, 2023

You Only Need One Thing One Click: Self-Training for Weakly Supervised 3D Scene Understanding
Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu
Self Training Good Thing 3D Scene Understanding Label Propagation 3D Instance Segmentation Sparse Label Accidental Click Pseudo Labeled Image Weakly Supervised 3D

September 15, 2022

Active Self-Training for Weakly Supervised 3D Scene Semantic Segmentation
Gengxin Liu, Oliver van Kaick, Hui Huang, Ruizhen Hu
Pseudo Label Self Training Semantic Segmentation Network Scene Segmentation Supervised Segmentation Weakly Supervised 3D

Weakly Supervised 3D

Papers

3D Weakly Supervised Semantic Segmentation with 2D Vision-Language Guidance

Contrastive Gaussian Clustering: Weakly Supervised 3D Scene Segmentation

Enhancing Weakly Supervised 3D Medical Image Segmentation through Probabilistic-aware Learning

Multi-modality Affinity Inference for Weakly Supervised 3D Semantic Segmentation

A Review and A Robust Framework of Data-Efficient 3D Scene Parsing with Traditional/Learned 3D Descriptors

All Points Matter: Entropy-Regularized Distribution Alignment for Weakly-supervised 3D Segmentation

You Only Need One Thing One Click: Self-Training for Weakly Supervised 3D Scene Understanding

Active Self-Training for Weakly Supervised 3D Scene Semantic Segmentation