DeepCorrect: Correcting DNN Models Against Image Distortions
DeepCorrect: Correcting DNN Models Against Image Distortions
复制标题
DOI：
10.1109/tip.2019.2924172
复制
发表时间：
2019-12-01
影响因子：
10.6
通讯作者：
Karam, Lina J.
中科院分区：
文献类型：
--
作者：
Borkar, Tejas S.;Karam, Lina J.
近年来，深度神经网络（DNN）的广泛使用促进了图像分类和对象识别等计算机视觉任务的性能大幅提高。在大多数现实的计算机视觉应用中，输入图像在图像获取或传输期间经历某种形式的图像失真，例如模糊和加性噪声。在原始图像上训练的深度网络在测试这种失真时表现不佳。在本文中，我们评估了高斯模糊和加性噪声等图像失真对预训练卷积滤波器激活的影响。我们提出了一个度量来识别最易受噪声影响的卷积滤波器，并按照校正后分类精度的最高增益对它们进行排名。在我们提出的称为DeepCorrect的方法中，我们在这些排名滤波器的输出端应用具有剩余连接的卷积层的小堆栈，并训练它们以纠正受失真影响最严重的滤波器激活，同时保持网络中其余的预训练滤波器输出不变。性能结果表明，将DeepCorrect模型应用于常见的视觉任务，如图像分类（ImageNet），对象识别（Caltech-101，Caltech-256）和场景分类（SUN-397），显着提高了DNN对失真图像的鲁棒性，并优于其他替代方法。
In recent years, the widespread use of deep neural networks (DNNs) has facilitated great improvements in performance for computer vision tasks like image classification and object recognition. In most realistic computer vision applications, an input image undergoes some form of image distortion such as blur and additive noise during image acquisition or transmission. Deep networks trained on pristine images perform poorly when tested on such distortions. In this paper, we evaluate the effect of image distortions like Gaussian blur and additive noise on the activations of pre-trained convolutional filters. We propose a metric to identify the most noise susceptible convolutional filters and rank them in order of the highest gain in classification accuracy upon correction. In our proposed approach called DeepCorrect, we apply small stacks of convolutional layers with residual connections at the output of these ranked filters and train them to correct the worst distortion affected filter activations, while leaving the rest of the pre-trained filter outputs in the network unchanged. Performance results show that applying DeepCorrect models for common vision tasks like image classification (ImageNet), object recognition (Caltech-101, Caltech-256), and scene classification (SUN-397), significantly improves the robustness of DNNs against distorted images and outperforms other alternative approaches.