CT-SRCNN: Cascade Trained and Trimmed Deep Convolutional Neural Networks for Image Super Resolution
This work addresses the need for efficient and accurate super-resolution models, which is incremental as it builds on existing deep CNN methods with optimization techniques.
The authors tackled the problem of improving both accuracy and efficiency in image super-resolution by proposing cascade training and trimming methods for deep CNNs, achieving state-of-the-art accuracy and over 4 times faster speed compared to existing networks.
We propose methodologies to train highly accurate and efficient deep convolutional neural networks (CNNs) for image super resolution (SR). A cascade training approach to deep learning is proposed to improve the accuracy of the neural networks while gradually increasing the number of network layers. Next, we explore how to improve the SR efficiency by making the network slimmer. Two methodologies, the one-shot trimming and the cascade trimming, are proposed. With the cascade trimming, the network's size is gradually reduced layer by layer, without significant loss on its discriminative ability. Experiments on benchmark image datasets show that our proposed SR network achieves the state-of-the-art super resolution accuracy, while being more than 4 times faster compared to existing deep super resolution networks.