How deep learning is empowering semantic segmentation: Traditional and deep learning techniques for semantic...

Area of Science:

• Computer Vision
• Machine Learning
• Image Processing

Background:

• Semantic segmentation extracts pixel-level information for detailed data analysis.
• Traditional methods relied on unsupervised learning or conventional image processing.
• Advancements have led to more efficient and accurate segmentation techniques.

Purpose of the Study:

• To comprehensively review supervised and unsupervised learning algorithms for semantic segmentation.
• To analyze the impact and advancements of deep learning in semantic segmentation.
• To synthesize findings from approximately 120 research papers in the field.

Main Methods:

• Systematic literature review of semantic segmentation algorithms.
• Analysis of both traditional and deep learning-based approaches.
• Comparative study of various supervised and unsupervised learning techniques.

Main Results:

• Deep learning methods offer significant improvements in semantic segmentation efficiency and accuracy.
• A wide range of algorithms, from basic to advanced, have been developed.
• Pixel-based classification is central to semantic segmentation's detailed data extraction.

Conclusions:

• Deep learning is pivotal in overcoming critical challenges in semantic segmentation.
• The reviewed algorithms demonstrate the evolution towards more effective image analysis.
• Further research continues to refine deep learning applications for enhanced semantic segmentation.