Related Experiment Video
Updated: Dec 5, 2025
05:12
Robotized Testing of Camera Positions to Determine Ideal Configuration for Stereo 3D Visualization of Open-Heart Surgery
Published on: August 12, 2021
2.4K
A Survey on Deep Learning Techniques for Stereo-Based Depth Estimation
IEEE Transactions on Pattern Analysis and Machine Intelligence
|October 20, 2020
Summary
Deep learning significantly improves stereo matching for depth estimation from RGB images, overcoming limitations of traditional methods. This survey covers recent advancements and future directions in this rapidly evolving field.
Area of Science:
- Computer Vision
- Machine Learning
- 3D Graphics
Background:
- Estimating depth from RGB images is a challenging, ill-posed problem.
- Stereo matching, inspired by human vision, is a common approach.
- Traditional stereo methods struggle with texture, uniform regions, and occlusions.
Purpose of the Study:
- To provide a comprehensive survey of deep learning for stereo-based depth estimation.
- To summarize common deep learning pipelines, their benefits, and limitations.
- To discuss future research directions in this domain.
Main Methods:
- Reviewing over 150 papers published between 2014 and 2019 on deep learning for stereo depth estimation.
- Analyzing and categorizing various deep learning architectures and techniques.
- Comparing performance and identifying challenges of different methods.
Main Results:
- Deep learning methods show significant performance improvements over traditional techniques.
- These advancements enable new applications in autonomous driving and augmented reality.
- The field has rapidly grown, with continuous innovation.
Conclusions:
- Deep learning has revolutionized stereo-based depth estimation.
- Further research is needed to address remaining challenges and explore future potential.
- The survey provides a valuable overview for researchers and practitioners.
