# Face Detection with Feature Pyramids and Landmarks

arXiv: https://arxiv.org/abs/1912.00596 · HTML: https://mlsanity.com/paper/1912.00596/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 Dec 2019.

## Built on by (most cited first, 5 of 5)

- [LOTR: Face Landmark Localization Using Localization Transformer](https://mlsanity.com/paper/2109.10057/) (2021)
- [Safe Fakes: Evaluating Face Anonymizers for Face Detectors](https://mlsanity.com/paper/2104.11721/) (2021)
- [Sub-pixel face landmarks using heatmaps and a bag of tricks](https://mlsanity.com/paper/2103.03059/) (2021)
- [Development of a face mask detection pipeline for mask-wearing monitoring in the era of the COVID-19 pandemic: A modular approach](https://mlsanity.com/paper/2112.15031/) (2021)
- [Fully Quantized Always-on Face Detector Considering Mobile Image Sensors](https://mlsanity.com/paper/2311.01001/) (2023)

## Abstract (arXiv)

Accurate face detection and facial landmark localization are crucial to any face recognition system. We present a series of three single-stage RCNNs with different sized backbones (MobileNetV2-25, MobileNetV2-100, and ResNet101) and a six-layer feature pyramid trained exclusively on the WIDER FACE dataset. We compare the face detection and landmark accuracies using eight context module architectures, four proposed by previous research and four modified versions. We find no evidence that any of the proposed architectures significantly overperform and postulate that the random initialization of the additional layers is at least of equal importance. To show this we present a model that achieves near state-of-the-art performance on WIDER FACE and also provides high accuracy landmarks with a simple context module. We also present results using MobileNetV2 backbones, which achieve over 90% average precision on the WIDER FACE hard validation set while being able to run in real-time. By comparing to other authors, we show that our models exceed the state-of-the-art for similar-sized RCNNs and match the performance of much heavier networks.