# Face Hallucination by Attentive Sequence Optimization with Reinforcement Learning

arXiv: https://arxiv.org/abs/1905.01509 · HTML: https://mlsanity.com/paper/1905.01509/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 4 May 2019.

## Built on by (most cited first, 8 of 8)

- [Deepfake Generation and Detection: A Benchmark and Survey](https://mlsanity.com/paper/2403.17881/) (2024)
- [RestoreFormer++: Towards Real-World Blind Face Restoration from Undegraded Key-Value Pairs](https://mlsanity.com/paper/2308.07228/) (2023)
- [Deep Learning-based Face Super-Resolution: A Survey](https://mlsanity.com/paper/2101.03749/) (2021)
- [Improving Face-Based Age Estimation with Attention-Based Dynamic Patch Fusion](https://mlsanity.com/paper/2112.10167/) (2021)
- [Criteria Comparative Learning for Real-scene Image Super-Resolution](https://mlsanity.com/paper/2207.12767/) (2022)
- [MO-MIX: Multi-Objective Multi-Agent Cooperative Decision-Making With Deep Reinforcement Learning](https://mlsanity.com/paper/2603.00730/) (2026)
- [GTAE: Graph-Transformer based Auto-Encoders for Linguistic-Constrained Text Style Transfer](https://mlsanity.com/paper/2102.00769/) (2021)
- [Super-Resolving Face Image by Facial Parsing Information](https://mlsanity.com/paper/2304.02923/) (2023)

## Abstract (arXiv)

Face hallucination is a domain-specific super-resolution problem that aims to generate a high-resolution (HR) face image from a low-resolution~(LR) input. In contrast to the existing patch-wise super-resolution models that divide a face image into regular patches and independently apply LR to HR mapping to each patch, we implement deep reinforcement learning and develop a novel attention-aware face hallucination (Attention-FH) framework, which recurrently learns to attend a sequence of patches and performs facial part enhancement by fully exploiting the global interdependency of the image. Specifically, our proposed framework incorporates two components: a recurrent policy network for dynamically specifying a new attended region at each time step based on the status of the super-resolved image and the past attended region sequence, and a local enhancement network for selected patch hallucination and global state updating. The Attention-FH model jointly learns the recurrent policy network and local enhancement network through maximizing a long-term reward that reflects the hallucination result with respect to the whole HR image. Extensive experiments demonstrate that our Attention-FH significantly outperforms the state-of-the-art methods on in-the-wild face images with large pose and illumination variations.