Deep Learning with Inaccurate Training Data for Image Restoration
This tackles the issue of inaccurate training data for image restoration, which is common in applications where real paired data is hard to obtain, though it appears incremental as it builds on existing synthetic training approaches.
The paper addresses the problem of poor performance of deep convolutional neural networks (DCNNs) on real-world image restoration tasks when trained on synthetic data due to generalization errors, and proposes a new training method that compensates for these errors to a large extent.
In many applications of deep learning, particularly those in image restoration, it is either very difficult, prohibitively expensive, or outright impossible to obtain paired training data precisely as in the real world. In such cases, one is forced to use synthesized paired data to train the deep convolutional neural network (DCNN). However, due to the unavoidable generalization error in statistical learning, the synthetically trained DCNN often performs poorly on real world data. To overcome this problem, we propose a new general training method that can compensate for, to a large extent, the generalization errors of synthetically trained DCNNs.