# Going Deeper Into Face Detection: A Survey

**Authors:** Shervin Minaee, Ping Luo, Zhe Lin, Kevin Bowyer  
**arXiv:** [2103.14983](https://arxiv.org/abs/2103.14983)  
**Date:** 2021-03-27  
**Summary page:** https://www.summarizepaper.com/en/paper/2103.14983v1/

## Key points

- Comprehensive survey of over fifty recent deep-learning-based face detection methods
- Single Shot Detector Models
- Feature Pyramid Network Based Models
- Cascade-CNN Based Models
- R-CNN and Faster-RCNN Based Models
- Discussion on various aspects including training data, network architectures, loss functions, and training strategies
- Overview of popular Deep Neural Network (DNN) architectures like AlexNet, VGGNet, and ResNet
- Review of significant state-of-the-art deep learning-based face detection models
- Summary of popular benchmarks for face detection in terms of size and characteristics
- Listing metrics for evaluating deep-learning-based face detection models and their performance on datasets
- Discussion on challenges and opportunities in deep learning-based face detection.

## Summary

This paper provides a comprehensive survey of over fifty recent deep-learning-based face detection methods. These methods are categorized based on their main technical contributions and grouped into categories such as Single Shot Detector Models, Feature Pyramid Network Based Models, Cascade-CNN Based Models, R-CNN and Faster-RCNN Based Models, and others. The paper delves into various aspects of these methods including training data, network architectures, loss functions, training strategies, and key contributions. The survey begins with an overview of popular Deep Neural Network (DNN) architectures commonly used in modern face detection algorithms such as AlexNet, VGGNet, and ResNet. It then reviews significant state-of-the-art deep learning-based face detection models and their technical advancements. Popular benchmarks for face detection are summarized in terms of size and characteristics while metrics for evaluating deep-learning-based face detection models are listed along with the performance of models on these datasets. Additionally, the paper discusses challenges and opportunities in deep learning-based face detection before presenting conclusions. Overall, this survey serves as a valuable resource for researchers and practitioners in the field of facial recognition systems by providing insights into the latest advancements in deep learning techniques for accurate face detection.