Awsome Delineation

Awesome Delineation is a comprehensive resource hub focused on topology-preserving delineation methods, which are essential for accurately extracting connected structures like road networks, blood vessels, and biological tissues from images. It aggregates decades of research to help practitioners understand and implement algorithms that maintain structural connectivity—a critical requirement in fields like remote sensing, medical imaging, and autonomous systems.

Key Features

• Academic Paper Collection — Curated list of seminal and recent papers on delineation, skeletonization, and topology-preserving segmentation from major conferences and journals.
• Dataset References — Links to key datasets such as SpaceNet and DeepGlobe used for training and evaluating road extraction and remote sensing models.
• Evaluation Metrics — Documentation of metrics like APLS (Average Path Length Similarity) and topology-aware loss functions for assessing reconstruction quality.
• Textbook Resources — Foundational textbooks on computational topology and digital topology to provide theoretical background.
• Multi-Domain Coverage — Resources span computer vision (aerial imagery, road extraction), medical imaging (vessel segmentation), and biological structure analysis.

Philosophy

The project emphasizes that preserving topological correctness—such as connectivity, loops, and network structure—is as important as pixel-wise accuracy in delineation tasks, especially for real-world applications where structural integrity matters.