Distortion Robust Image Classification with Deep Convolutional Neural Network based on Discrete Cosine Transform

11/14/2018
by   Md Tahmid Hossain, et al.
8

State of the art CNN models for image classification are found to be highly vulnerable to image quality degradation. It is observed that even a small amount of distortion introduced in an image in the form of noise or blur severely hampers the performance of these CNN architectures. Most of the work in the literature strive to mitigate this problem simply by fine-tuning a pre-trained model on mutually exclusive or union set of distorted training data. This iterative fine-tuning process with all possible types of distortion is exhaustive and struggles to handle unseen distortions. In this work, we propose DCT-Net, a Discrete Cosine Transform based module integrated into a deep network which is built on top of VGG16 vgg1. The proposed DCT module operates during training and discards input information based on DCT coefficients which represent the contribution of sampling frequencies. We show that this approach enables the network to be trained at one go without having to generate training data with different type of expected distortions. We also extend the idea of traditional dropout and present a training adaptive version of the same. During tests, we introduce Gaussian blur, motion blur, salt and pepper noise, Gaussian white noise and speckle noise to CIFAR10, CIFAR-100 cifar1 and ImageNet imagenet1 dataset. We evaluate our deep network on these benchmark databases and show that it not only generalizes well to a variety of image distortions but also outperforms sate-of-the-art.

READ FULL TEXT

page 1

page 2

page 4

page 5

research
11/14/2018

Distortion Robust Image Classification using Deep Convolutional Neural Network with Discrete Cosine Transform

Convolutional Neural Network is good at image classification. However, i...
research
01/08/2017

On Classification of Distorted Images with Deep Convolutional Neural Networks

Image blur and image noise are common distortions during image acquisiti...
research
05/05/2017

DeepCorrect: Correcting DNN models against Image Distortions

In recent years, the widespread use of deep neural networks (DNNs) has f...
research
11/17/2016

Examining the Impact of Blur on Recognition by Convolutional Networks

State-of-the-art algorithms for many semantic visual tasks are based on ...
research
12/02/2019

Training the Convolutional Neural Network with Statistical Dependence of the Response on the Input Data Distortion

The paper proposes an approach to training a convolutional neural networ...
research
04/05/2020

Comparative Analysis of Multiple Deep CNN Models for Waste Classification

Waste is a wealth in a wrong place. Our research focuses on analyzing po...
research
10/03/2012

Blurred Image Classification based on Adaptive Dictionary

Two types of framework for blurred image classification based on adaptiv...

Please sign up or login with your details

Forgot password? Click here to reset