# A Survey of Modern Object Detection Literature using Deep Learning

**Authors:** Karanbir Singh Chahal, Kuntal Dey  
**arXiv:** [1808.07256](https://arxiv.org/abs/1808.07256)  
**Date:** 2018-08-22  
**Summary page:** https://www.summarizepaper.com/en/paper/1808.07256v1/

## Key points

- The paper provides a comprehensive overview of object detection algorithms using deep learning techniques
- Object detection involves identifying, localizing, and classifying objects in images
- The survey analyzes different aspects of modern object detection algorithms, including algorithms themselves, quality metrics, speed/size trade-offs, and training methodologies
- The authors focus on the Single Shot MultiBox Detector (SSD) and Faster Region-based Convolutional Neural Network (Faster R-CNN) algorithms
- Portable and efficient detectors for low-powered devices are addressed by exploring lightweight convolutional base architectures
- Strengths and weaknesses of each detector are thoroughly reviewed
- The paper presents an up-to-date understanding of the state-of-the-art in object detection using deep learning techniques
- It serves as a comprehensive resource for researchers and practitioners interested in modern object detection literature

## Summary

The paper titled "A Survey of Modern Object Detection Literature using Deep Learning" by Karanbir Singh Chahal and Kuntal Dey provides a comprehensive overview of object detection algorithms that utilize deep learning techniques. Object detection is a crucial task in computer vision, as it involves identifying objects in images and accurately localizing and classifying them. This survey aims to analyze various aspects of modern object detection algorithms, including different algorithms themselves, quality metrics, trade-offs between speed and size, and training methodologies. The authors specifically focus on two types of object detection algorithms: the Single Shot MultiBox Detector (SSD) class of single-step detectors and the Faster Region-based Convolutional Neural Network (Faster R-CNN) class of two-step detectors. These algorithms have gained significant attention due to their effectiveness in detecting objects in images. Additionally, the paper addresses the challenge of constructing detectors that are portable and efficient on low-powered devices. To tackle this issue, the authors explore new lightweight convolutional base architectures that can enable fast inference on resource-constrained devices. Throughout the survey, the strengths and weaknesses of each detector are thoroughly reviewed, providing valuable insights into their performance characteristics. By analyzing these factors, the paper presents an up-to-date understanding of the state-of-the-art in object detection using deep learning techniques. Overall, this survey serves as a comprehensive resource for researchers and practitioners interested in understanding modern object detection literature. It not only covers various algorithms but also delves into important considerations such as quality metrics, speed/size trade-offs, training methodologies, and lightweight architectures.