Recently, two papers, “Learning a Unified Classifier Incrementally via Rebalancing” from Saihui Hou and “Meta-SR: A Magnification-Arbitrary Network for Super-Resolution” from Xuecai Hu, are accepted by IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2019), which is one of the top conference for computer vision.
1. Learning a Unified Classifier Incrementally via Rebalancing
Abstract: Conventionally, deep neural networks are trained offline, relying on a large dataset prepared in advance. This paradigm is often challenged in real-world applications, e.g. online services that involve continuous streams of incoming data. Recently, incremental learning receives increasing attention, and is considered as a promising solution to the practical challenges mentioned above. However, it has been observed that incremental learning is subject to
a fundamental difficulty – catastrophic forgetting, namely adapting a model to new data often results in severe performance degradation on previous tasks or classes. Our study reveals that the imbalance between previous and new data is a crucial cause to this problem. In this work, we develop a new framework for incrementally learning a unified classifier, i.e. a classifier that treats both old and new classes uniformly. Specifically, we incorporate three components,
cosine normalization, less-forget constraint, and inter-class separation, to mitigate the adverse effects of the imbalance. Experiments show that the proposed method can effectively rebalance the training process, thus obtaining superior performance compared to the existing methods. On CIFAR100 and ImageNet, our method can reduce the classification errors by more than 6% and 13% respectively, under the incremental setting of 10 phases.
2. Meta-SR: A Magnification-Arbitrary Network for Super-Resolution
Abstract: Recent research on super-resolution has achieved great success due to the development of deep convolutional neural networks (DCNNs). However, super-resolution of arbitrary scale factor has been ignored for a long time. Most previous researchers regard super-resolution of different scale factors as independent tasks. They train a specific model for each scale factor which is inefficient in computing, and prior work only take the super-resolution of several integer scale factors into consideration. In this work, we propose a novel method called Meta-SR to firstly solve super-resolution of arbitrary scale factor (including noninteger scale factors) with a single model. In our Meta-SR, the Meta-Upscale Module is proposed to replace the traditional upscale module. For arbitrary scale factor, the MetaUpscale Module dynamically predicts the weights of the upscale filters by taking the scale factor as input and use these weights to generate the HR image of arbitrary size. For any low-resolution image, our Meta-SR can continuously zoom in it with arbitrary scale factor by only using a single model. We evaluated the proposed method through extensive experiments on widely used benchmark datasets on single image super-resolution. The experimental results show the superiority of our Meta-Upscale.