# Deep Learning vs. Traditional Computer Vision

**Authors:** Niall O' Mahony, Sean Campbell, Anderson Carvalho, Suman Harapanahalli, Gustavo Velasco-Hernandez, Lenka Krpalkova, Daniel Riordan, Joseph Walsh  
**arXiv:** [1910.13796](https://arxiv.org/abs/1910.13796)  
**Date:** 2019-10-30  
**Summary page:** https://www.summarizepaper.com/en/paper/1910.13796v1/

## Key points

- Deep learning has revolutionized digital image processing but traditional computer vision techniques are still relevant
- The paper analyzes the benefits and drawbacks of both approaches
- Maintaining knowledge of classical computer vision is discussed
- Hybrid methodologies combining deep learning and traditional computer vision have shown improved performance
- Panoramic vision and 3D vision are highlighted as areas where combining the two approaches is popular
- Overall, the paper provides a comprehensive analysis of the ongoing debate between deep learning and traditional computer vision techniques in digital image processing

## Summary

This paper titled "Deep Learning vs. Traditional Computer Vision" by Niall O' Mahony, Sean Campbell, Anderson Carvalho, Suman Harapanahalli, Gustavo Velasco-Hernandez, Lenka Krpalkova, Daniel Riordan and Joseph Walsh explores the ongoing debate between deep learning and traditional computer vision techniques in the field of digital image processing. While deep learning has revolutionized the domain by pushing the limits of what was previously possible, this does not render traditional computer vision techniques obsolete. The authors aim to analyze both approaches' benefits and drawbacks while promoting a discussion on whether knowledge of classical computer vision should be maintained. The paper also delves into how these two sides of computer vision can be combined to enhance performance and tackle problems that are not well-suited for deep learning models. Several recent hybrid methodologies have been reviewed which have demonstrated improved computer vision performance. Emerging domains such as panoramic vision and 3D vision are highlighted where combining traditional computer vision techniques with deep learning has been particularly popular due to the fact that deep learning models have not yet been fully optimized for these areas. Overall, this paper provides a comprehensive analysis of the ongoing debate between deep learning and traditional computer vision techniques in digital image processing. It explores both approaches' benefits and drawbacks while emphasizing the potential for combining them to achieve enhanced performance in various domains.