Online learning for automatic segmentation of 3D data

We propose a method to perform automatic segmentation of 3D scenes based on a standard classifier, whose learning model is continuously improved by means of new samples, and a grouping stage, that enforces local consistency among classified labels. The new samples are automatically delivered to the system by a feedback loop based on a feature selection approach that exploits the outcome of the grouping stage. By experimental results on several datasets we demonstrate that the proposed online learning paradigm is effective in increasing the accuracy of the whole 3D segmentation thanks to the improvement of the learning model of the classifier by means of newly acquired, unsupervised data.

Authors

• Federico Tombari
• Luigi Di Stefano
• Simone Giardino

Keywords

• Three dimensional displays
• Training
• Feature extraction
• Support vector machines
• Shape
• Image color analysis
• Solid modeling
• Online Learning
• 3D Data
• Learning Models
• 3D Segmentation
• Feature Selection Approach
• Standard Classifier
• Training Set
• Semantic Segmentation
• Categorical Model
• Class Position
• Original Approach
• Training Characteristics
• Robotic Applications
• 3D Features
• Markov Random Field
• Final Segmentation
• Conditional Random Field
• Computer Vision Applications
• Intra-class Variance
• Removal Procedure
• New York City
• Pairwise Terms
• 3D Sensor
• First-in-first-out
• Unsupervised Way
• Online Learning Methods
• Class Assignment
• Components Of The Graph
• Robot Vision
• Nodes In The Graph