[论文解读] Recurrent Exposure Generation for Low-Light Face Detection

Jinxiu Liang, Jingwen Wang+5 more|arXiv (Cornell University)|Jul 21, 2020

Video Surveillance and Tracking Methods参考文献 58被引用 11

一句话总结

本文提出 REGDet，一种用于低光照人脸检测的端到端检测增强框架，通过双向 ConvGRU 构建的循环曝光生成（REG）模块，从单张低光照输入中合成多张伪曝光图像，再由多曝光检测（MED）模块进行特征融合。该方法在 DARK FACE 基准测试中达到最先进性能，仅带来可忽略的参数增加。

ABSTRACT

Face detection from low-light images is challenging due to limited photos and inevitable noise, which, to make the task even harder, are often spatially unevenly distributed. A natural solution is to borrow the idea from multi-exposure, which captures multiple shots to obtain well-exposed images under challenging conditions. High-quality implementation/approximation of multi-exposure from a single image is however nontrivial. Fortunately, as shown in this paper, neither is such high-quality necessary since our task is face detection rather than image enhancement. Specifically, we propose a novel Recurrent Exposure Generation (REG) module and couple it seamlessly with a Multi-Exposure Detection (MED) module, and thus significantly improve face detection performance by effectively inhibiting non-uniform illumination and noise issues. REG produces progressively and efficiently intermediate images corresponding to various exposure settings, and such pseudo-exposures are then fused by MED to detect faces across different lighting conditions. The proposed method, named REGDet, is the first `detection-with-enhancement' framework for low-light face detection. It not only encourages rich interaction and feature fusion across different illumination levels, but also enables effective end-to-end learning of the REG component to be better tailored for face detection. Moreover, as clearly shown in our experiments, REG can be flexibly coupled with different face detectors without extra low/normal-light image pairs for training. We tested REGDet on the DARK FACE low-light face benchmark with thorough ablation study, where REGDet outperforms previous state-of-the-arts by a significant margin, with only negligible extra parameters.

研究动机与目标

• 为解决低光照人脸检测中的挑战，即光照不足与噪声会降低特征提取与检测性能。
• 克服独立图像增强方法的局限性，后者可能损害检测所需的判别性特征。
• 开发一种联合学习框架，整合曝光生成与人脸检测，以提升在非均匀光照条件下的鲁棒性。
• 实现面向人脸检测量身定制的、有效的端到端曝光生成训练，而非通用图像质量优化。

提出的方法

• REG 模块利用双向 ConvGRUs 从单张低光照输入中生成一系列逐步曝光的伪图像，通过循环记忆建模曝光变化。
• REG 模块通过伪监督预训练方案进行训练，无需真实配对曝光图像，即可引导其生成多样化的曝光水平。
• MED 模块采用多尺度特征聚合策略，融合多张伪曝光图像的特征，以提升检测鲁棒性。
• REG 与 MED 模块按顺序连接，并在端到端训练中联合优化，实现跨曝光层级的特征交互。
• 该框架兼容多种基础检测器（如 DSFD、PyramidBox），且训练无需配对的低光照或正常光照图像。
• 该架构轻量化，仅引入少量额外参数，却实现显著的性能提升。

实验结果

研究问题

• RQ1 循环曝光生成模块能否有效从单张低光照图像中合成伪曝光图像，以提升人脸检测性能？
• RQ2 曝光生成与检测的端到端联合学习是否优于先预处理增强再进行检测的方法？
• RQ3 生成的曝光阶段数量如何影响检测性能与计算成本？
• RQ4 REG 模块的伪监督预训练是否能提升特征多样性与检测准确率？
• RQ5 REGDet 框架在不同基础人脸检测器上的泛化能力如何？

主要发现

• 当以 PyramidBox 为基础检测器时，REGDet 在 DARK FACE 基准测试中达到 77.69% 的 mAP，显著优于此前最先进方法。
• 消融实验表明，REG 模块中使用 T=4 个阶段时性能最佳，T=6 时无进一步增益且计算成本更高。
• REG 模块的伪监督预训练带来 1.33% 的绝对 mAP 提升（从 76.36% 提升至 77.69%），证明其在学习互补曝光特征方面的有效性。
• 将 REG 替换为非循环模块（如 BEG、CEG 或 SEG）导致性能下降或不稳定，凸显循环记忆在建模曝光演变过程中的重要性。
• 该方法在不同基础检测器（DSFD、PyramidBox、S3FD）上均保持优异性能，证明其可扩展性与兼容性。
• 即使仅增加少量额外参数，REGDet 仍实现显著的性能提升，证实其高效性与实用性。