[论文解读] Perceptual Generative Adversarial Networks for Small Object Detection

Jianan Li, Xiaodan Liang+4 more|arXiv (Cornell University)|Jun 16, 2017

Advanced Neural Network Applications参考文献 39被引用 88

一句话总结

论文提出 Perceptual GAN，它学习把小物体特征转化为超分辨的特征，使之看起来更像大物体的特征，以提升检测，在交通标志和行人数据上得到验证。

ABSTRACT

Detecting small objects is notoriously challenging due to their low resolution and noisy representation. Existing object detection pipelines usually detect small objects through learning representations of all the objects at multiple scales. However, the performance gain of such ad hoc architectures is usually limited to pay off the computational cost. In this work, we address the small object detection problem by developing a single architecture that internally lifts representations of small objects to "super-resolved" ones, achieving similar characteristics as large objects and thus more discriminative for detection. For this purpose, we propose a new Perceptual Generative Adversarial Network (Perceptual GAN) model that improves small object detection through narrowing representation difference of small objects from the large ones. Specifically, its generator learns to transfer perceived poor representations of the small objects to super-resolved ones that are similar enough to real large objects to fool a competing discriminator. Meanwhile its discriminator competes with the generator to identify the generated representation and imposes an additional perceptual requirement - generated representations of small objects must be beneficial for detection purpose - on the generator. Extensive evaluations on the challenging Tsinghua-Tencent 100K and the Caltech benchmark well demonstrate the superiority of Perceptual GAN in detecting small objects, including traffic signs and pedestrians, over well-established state-of-the-arts.

研究动机与目标

• 阐明由于低分辨率表示导致的小目标检测挑战。
• 提出一个基于 GAN 的框架，为小目标生成超分辨表示以辅助检测。
• 利用条件生成器和感知判别器使小目标特征与大目标特征对齐。
• 在交通标志和行人基准数据集上显示相较于最先进方法的改进。

提出的方法

• 引入一个学习来自低层特征残差的生成器，以产生小对象的超分辨表示。
• 使用具有对抗和感知分支的判别器来监督生成，确保有利于检测的表示。
• 交替步骤训练生成器和判别器，以提升小目标检测性能。
• 应用 RoI 池化和残差连接，将生成的特征整合到检测管线。

实验结果

研究问题

• RQ1 感知 GAN 能否将小目标表示转换为大目标样的表示以提高检测精度？
• RQ2 对抗和感知监督是否提升小目标检测的生成特征质量？
• RQ3 使用不同的底层特征作为生成器输入对小目标检测性能的影响？
• RQ4 该方法是否超出 Traffic Sign 和 Pedestrian 数据集具有普适性？
• RQ5 相较于端到端基线，替代训练是否提升小对象性能？

主要发现

• 与现有方法相比，Perceptual GAN 在交通标志和行人上的小目标检测的召回率与准确度有所提升。
• 使用较低层特征（Conv1）作为生成器输入比高层特征带来更好的小目标结果。
• 交替优化（生成器-判别器）优于端到端基线训练。
• 方法在小对象上取得显著性能提升，同时保持总体目标检测性能。
• 泛化实验表明在 VOC 上小实例检测有所改善，基线超越标准方法。