恢复关键内容：联合恢复与识别的经验教训
Restore What Matters: Lessons from Joint Restoration and Recognition
- The University of Texas at Austin（德克萨斯大学奥斯汀分校）
- Purdue University（普渡大学）
- Michigan State University（密歇根州立大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本文提出面向识别的联合恢复范式JR²，仅恢复任务所需内容，结合物理模拟、神经科学和端到端识别损失，在IARPA-BRIAR上提升性能并降低计算成本。
AI中文摘要：
识别流水线通常采用先恢复后识别的流程，然而数十年的经验表明，生成视觉上美观的图像很少能转化为识别性能的提升。我们提出了一种面向识别的联合恢复（JR$^2$）范式：仅恢复下游任务真正需要的内容，由任务信号决定恢复的位置、程度以及是否必要。JR$^2$ 建立在三个支柱之上：（i）物理学，采用光学精确的湍流模拟（可扩展到模糊和噪声），将恢复建立在真实图像形成过程的基础上；（ii）神经科学，借鉴选择性注意和神经可塑性，将模型能力引导至身份关键区域和帧，同时跳过已干净的输入；（iii）视觉与学习，通过恢复和对齐将识别损失端到端耦合，使低级编辑最大化高级身份稳定性。在IARPA-BRIAR上的评估显示出一致的改进（例如，TAR@0.01% FAR +0.6；FNIR@1% FPIR -2.5），而质量门控跳过了约70%的干净帧，降低了成本。消融实验证实物理先验增强了真实性，联合训练防止了灾难性遗忘，且在许多情况下选择性恢复已足够。我们得出结论：更美观的图像既非必要也非充分；恢复模块必须是任务驱动、选择性且具有物理感知的。提供了代码、预训练模型和集成配方。
英文摘要：
Recognition pipelines typically adopt a restore-then-recognize workflow, yet decades of experience show that generating visually pleasing images seldom translates to improved recognition. We propose a Joint Restoration-for-Recognition (JR$^2$) paradigm: restore only what downstream tasks truly require, with task signals dictating where, how much, and whether restoration is necessary. JR$^2$ rests on three pillars: (i) Physics, employing optics-accurate turbulence simulation, extensible to blur and noise, to ground restoration in real image formation; (ii) Neuroscience, drawing on selective attention and neuroplasticity to direct model capacity toward identity-critical regions and frames while bypassing already-clean inputs; and (iii) Vision & Learning, coupling recognition loss end-to-end through restoration and alignment so that low-level edits maximize high-level identity stability. Evaluations on IARPA-BRIAR show consistent improvements (e.g., TAR@0.01% FAR +0.6; FNIR@1% FPIR -2.5), while a quality gate skips ~70% of clean frames, reducing cost. Ablations confirm physics priors enhance realism, joint training prevents catastrophic forgetting, and selective restoration suffices in many cases. We conclude that better-looking images are neither necessary nor sufficient; restoration modules must be task-driven, selective, and physically aware. Code, pretrained models, and recipes are provided for integration.
