# Face Parsing via a Fully-Convolutional Continuous CRF Neural Network

arXiv: https://arxiv.org/abs/1708.03736 · HTML: https://mlsanity.com/paper/1708.03736/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Aug 2017.

## Built on by (most cited first, 7 of 7)

- [Decoupled Multi-task Learning with Cyclical Self-Regulation for Face Parsing](https://mlsanity.com/paper/2203.14448/) (2022)
- [RoI Tanh-polar Transformer Network for Face Parsing in the Wild](https://mlsanity.com/paper/2102.02717/) (2021)
- [DisFaceRep: Representation Disentanglement for Co-occurring Facial Components in Weakly Supervised Face Parsing](https://mlsanity.com/paper/2508.01250/) (2025)
- [CADA-GAN: Context-Aware GAN with Data Augmentation](https://mlsanity.com/paper/2301.08849/) (2023)
- [Occlusion-Aware Deep Convolutional Neural Network via Homogeneous Tanh-transforms for Face Parsing](https://mlsanity.com/paper/2308.15323/) (2023)
- [Generative Face Parsing Map Guided 3D Face Reconstruction Under Occluded Scenes](https://mlsanity.com/paper/2412.18920/) (2024)
- [Learning Contour-Guided 3D Face Reconstruction with Occlusions](https://mlsanity.com/paper/2503.12494/) (2025)

## Abstract (arXiv)

In this work, we address the face parsing task with a Fully-Convolutional continuous CRF Neural Network (FC-CNN) architecture. In contrast to previous face parsing methods that apply region-based subnetwork hundreds of times, our FC-CNN is fully convolutional with high segmentation accuracy. To achieve this goal, FC-CNN integrates three subnetworks, a unary network, a pairwise network and a continuous Conditional Random Field (C-CRF) network into a unified framework. The high-level semantic information and low-level details across different convolutional layers are captured by the convolutional and deconvolutional structures in the unary network. The semantic edge context is learnt by the pairwise network branch to construct pixel-wise affinity. Based on a differentiable superpixel pooling layer and a differentiable C-CRF layer, the unary network and pairwise network are combined via a novel continuous CRF network to achieve spatial consistency in both training and test procedure of a deep neural network. Comprehensive evaluations on LFW-PL and HELEN datasets demonstrate that FC-CNN achieves better performance over the other state-of-arts for accurate face labeling on challenging images.