Deepsegmentor: Multi-Task Crack and Road Network Segmentation

Deepsegmentor provides a PyTorch-based implementation of the DeepCrack and RoadNet architectures for infrastructure segmentation, updated by yhlleo in December 2024. The repository focuses on multi-task learning for crack detection, road extraction, and centerline identification.

Use Cases

• Training deep neural networks for crack-detection using binary segmentation masks
• Developing multi-task learning models for road-detection and centerline-detection
• Evaluating edge-detection algorithms on infrastructure imagery

Strengths

• Includes multi-task learning labels for simultaneous road and centerline detection
• Supports edge-detection specific labels for crack analysis
• Provides PyTorch implementations for DeepCrack and RoadNet architectures

Limitations

• Lack of explicit license information may restrict commercial usage
• Geographic origin of the infrastructure images is not specified

Provenance

Source

yhlleo (GitHub)

Freshness

Last updated December 2024.