Enhancing the Performance of Convolutional Neural Networks on Quality Degraded Datasets
This addresses a specific issue in image classification for applications dealing with real-world degraded data, but it is incremental as it builds on existing architectures.
The paper tackles the problem of convolutional neural networks performing poorly on quality-degraded datasets, such as those with noise or blur, by proposing a dual-channel model that improves overall performance compared to a single model.
Despite the appeal of deep neural networks that largely replace the traditional handmade filters, they still suffer from isolated cases that cannot be properly handled only by the training of convolutional filters. Abnormal factors, including real-world noise, blur, or other quality degradations, ruin the output of a neural network. These unexpected problems can produce critical complications, and it is surprising that there has only been minimal research into the effects of noise in the deep neural network model. Therefore, we present an exhaustive investigation into the effect of noise in image classification and suggest a generalized architecture of a dual-channel model to treat quality degraded input images. We compare the proposed dual-channel model with a simple single model and show it improves the overall performance of neural networks on various types of quality degraded input datasets.