Pre-Training Without Natural Images
Pre-Training Without Natural Images
复制标题
DOI：
10.1007/s11263-021-01555-8
复制
发表时间：
2021-01
影响因子：
19.5
通讯作者：
Hirokatsu Kataoka;Kazushige Okayasu;Asato Matsumoto;Eisuke Yamagata;Ryosuke Yamada;Nakamasa Inoue;Akio Nakamura;Y. Satoh
中科院分区：
文献类型：
--
作者：
Hirokatsu Kataoka;Kazushige Okayasu;Asato Matsumoto;Eisuke Yamagata;Ryosuke Yamada;Nakamasa Inoue;Akio Nakamura;Y. Satoh
有没有可能使用在没有任何自然图像的情况下预先训练的卷积神经网络来帮助理解自然图像？本文提出了一个新的概念，公式驱动的监督学习。我们自动生成图像模式和它们的类别标签分配分形，这是基于一个自然规律存在于真实的世界的背景知识。从理论上讲，在预训练阶段使用自动生成的图像而不是自然图像可以让我们生成无限规模的标记图像数据集。虽然使用所提出的分形数据库（FractalDB）（一种没有自然图像的数据库）预训练的模型在所有设置下都不一定优于使用人类注释数据集预训练的模型，但我们能够部分超过ImageNet/Places预训练模型的准确性。使用所提出的FractalDB的图像表示在卷积层和注意力的可视化中捕获了一个独特的特征。
Is it possible to use convolutional neural networks pre-trained without any natural images to assist natural image understanding? The paper proposes a novel concept, Formula-driven Supervised Learning. We automatically generate image patterns and their category labels by assigning fractals, which are based on a natural law existing in the background knowledge of the real world. Theoretically, the use of automatically generated images instead of natural images in the pre-training phase allows us to generate an infinite scale dataset of labeled images. Although the models pre-trained with the proposed Fractal DataBase (FractalDB), a database without natural images, does not necessarily outperform models pre-trained with human annotated datasets at all settings, we are able to partially surpass the accuracy of ImageNet/Places pre-trained models. The image representation with the proposed FractalDB captures a unique feature in the visualization of convolutional layers and attentions.