In recent years, coarse-to-fine networks have shown good performance in image deblurring. The image details recovered better by inputting multi-scale images, while the computation will be high. To design a efficient deblurring network, we use a new coarse-to-fine strategy for image deblurring. We use UNet as the backbone network, feeding one image of different scales at each layer of the network to obtain sharp images with better details. In addition, we have designed information supplement blocks that can effectively supplement the information of different scale images. Finally, to be able to obtain better information about the image, we introduce an attention mechanism. It is experimentally shown that our network performs well on different metrics.

The hash method or product quantization based on deep learning has achieved great success in image retrieval. But most deep hash methods are designed for supervised scenes. They only use semantic similarity information and ignore the underlying data structure. Moreover, a large amount of manual label information is expensive and time-consuming, which is not in line with the actual application scenario. In order to tackle this problem, we propose a novel quantization-based semi-supervised image retrieval network: Central Product Quantization Network (CPQN). We design a novel central similarity strategy to preserve the semantic similarity and underlying data structure in labeled data, and generalize it to unlabeled data through consistent regularization to tap the potential of unlabeled data. We also propose a novel semi-supervised loss algorithm to achieve effective hashing by reducing quantization noise and minimizing the empirical error of labeled data and the embedding error of unlabeled data. Experiments on public benchmark dataset clearly show that our proposed method is superior to the most advanced hash method.