Improving LiDAR Semantic Segmentation on Minority Classes and Generalization Capability for Autonomous Driving

Keyphrases

• Autonomous Driving 100%
• Autonomous Driving System 20%
• Class Class 20%
• Class Imbalance Issue 20%
• Convolution Operator 20%
• Cross-entropy 20%
• Dice Loss 20%
• Generalization Ability 100%
• Geometric Measurement 20%
• Large-scale Dataset 20%
• Large-scale Scene 20%
• LiDAR 40%
• LiDAR Data 20%
• LiDAR Semantic Segmentation 100%
• LiDAR Sensor 20%
• Loss Function 20%
• Majority Class 20%
• Majority-minority 20%
• Minority Class 100%
• MIoU 80%
• Motorcyclist 20%
• Object Detection 20%
• Perception Problem 20%
• Performance Gap 20%
• Semantic Navigation 20%
• Semantic Segmentation 40%
• Traffic Signs 20%
• UNet++ 20%

Computer Science

• Autonomous Driving 100%
• Class Imbalance 20%
• Convolution 20%
• Entropy 20%
• Image Segmentation 100%
• Majority Class 20%
• Minority Class 100%
• Object Detection 20%
• Performance Gap 20%
• U-Net 20%