One Class One Click: Quasi scene-level weakly supervised point cloud semantic segmentation with active learning

Keyphrases

• Unsupervised Learning 100%
• Active Learning 100%
• Point Cloud Segmentation 100%
• Class One 100%
• Competitive Results 40%
• F1 Score 40%
• Global Perspective 20%
• Model Prediction 20%
• Classification Task 20%
• Scene Classification 20%
• Learning Scheme 20%
• Global Features 20%
• Feature Embedding 20%
• Uncertainty Measure 20%
• LiDAR 20%
• Contextual Constraint 20%
• Publicly Available 20%
• Context-aware 20%
• Local Perspectives 20%
• LiDAR Dataset 20%
• Labeling Scheme 20%
• Labeling Cost 20%
• Large-scale Point Clouds 20%
• Supervisory Signal 20%
• Weakly Supervised Method 20%
• Data Annotation 20%
• Pseudo-labeling 20%
• Full Supervision 20%
• Weak Labels 20%
• Annotation Cost 20%
• Prediction Aggregation 20%
• Terrestrial LiDAR 20%
• Supervision Strategy 20%
• Labeling Strategy 20%
• Informative Sample 20%

Computer Science

• Annotation 100%
• Active Learning 100%
• Image Segmentation 100%
• Point Cloud 100%
• Global Feature 25%
• Baseline Model 25%
• Supervised Method 25%
• Labeling Scheme 25%
• Learning Scheme 25%
• Model Prediction 25%
• Classification Task 25%