Image restoration is a challenging ill-posed problem which also has been a long-standing issue. In the past few years, the convolution neural networks (CNNs) almost dominated the computer vision and had achieved considerable success in different levels of vision tasks including image restoration. However, recently the Swin Transformer-based model also shows impressive performance, even surpasses the CNN-based methods to become the state-of-the-art on high-level vision tasks. In this paper, we proposed a restoration model called SUNet which uses the Swin Transformer layer as our basic block and then is applied to UNet architecture for image denoising. The source code and pre-trained models are available at https://github.com/FanChiMao/SUNet.

SUNet: Swin Transformer UNet for Image Denoising

Image restoration is a challenging ill-posed problem which also has been a long-standing issue. In the past few years, the convolution neural networks (CNNs) almost dominated the computer vision and had achieved considerable success in different levels of vision tasks including image restoration. However, recently the Swin Transformer-based model also shows impressive performance, even surpasses the CNN-based methods to become the state-of-the-art on high-level vision tasks. In this paper, we proposed a restoration model called SUNet which uses the Swin Transformer layer as our basic block and then is applied to UNet architecture for image denoising. The source code and pre-trained models are available at https://github.com/FanChiMao/SUNet. 

/pdf/sunet-swin-transformer-unet-for-image-denoising-38ws01ea.pdf

In this paper, we study the problem of semantic part segmentation for animals. This is more challenging than standard object detection, object segmentation and pose estimation tasks because semantic parts of animals often have similar appearance and highly varying shapes. To tackle these challenges, we build a mixture of compositional models to represent the object boundary and the boundaries of semantic parts. And we incorporate edge, appearance, and semantic part cues into the compositional model. Given part-level segmentation annotation, we develop a novel algorithm to learn a mixture of compositional models under various poses and viewpoints for certain animal classes. Furthermore, a linear complexity algorithm is offered for efficient inference of the compositional model using dynamic programming. We evaluate our method for horse and cow using a newly annotated dataset on Pascal VOC 2010 which has pixelwise part labels. Experimental results demonstrate the effectiveness of our method.

Semantic Part Segmentation using Compositional Model combining Shape and Appearance

Low-Light Image Enhancement is a computer vision task which intensifies the dark images to appropriate brightness. It can also be seen as an ill-posed problem in image restoration domain. With the success of deep neural networks, the convolutional neural networks surpass the traditional algorithm-based methods and become the mainstream in the computer vision area. To advance the performance of enhancement algorithms, we propose an image enhancement network (HWMNet) based on an improved hierarchical model: M-Net+. Specifically, we use a half wavelet attention block on M-Net+ to enrich the features from wavelet domain. Furthermore, our HWMNet has competitive performance results on two image enhancement datasets in terms of quantitative metrics and visual quality. The source code and pretrained model are available at https://github.com/FanChiMao/HWMNet. 

/pdf/half-wavelet-attention-on-m-net-for-low-light-image-17xm50at.pdf

Half Wavelet Attention on M-Net+ for Low-Light Image Enhancement

Image restoration is a low-level vision task which is to restore degraded images to noise-free images. With the success of deep neural networks, the convolutional neural networks surpass the traditional restoration methods and become the mainstream in the computer vision area. To advance the performanceof denoising algorithms, we propose a blind real image denoising network (SRMNet) by employing a hierarchical architecture improved from U-Net. Specifically, we use a selective kernel with residual block on the hierarchical structure called M-Net to enrich the multi-scale semantic information. Furthermore, our SRMNet has competitive performance results on two synthetic and two real-world noisy datasets in terms of quantitative metrics and visual quality. The source code and pretrained model are available at https://github.com/TentativeGitHub/SRMNet. 

Selective Residual M-Net for Real Image Denoising

For the epidemic prevention of avian influenza, there exist lots of differences between ideality and reality. This is why the epidemic is usually out of control. One of the reasons is that many illegal waterfowl farms are built without government registration. In this work, we proposed a new method trying to directly locate waterfowl farms, including both registered and unregistered ones without the need of human labeling. This will not only save human labors, but also update the location and size information of waterfowl farms regularly due to the computing speed of computers. In this work, we proposed a new method for satellite image augmentation. The layers of the model we proposed are not deeper than the other deep neural network models. However, we show that using the existing simple U-Net combined with residual blocks has better performance than the other deep models in this task.

Locating Waterfowl Farms from Satellite Images with Parallel Residual U-Net Architecture

In this paper, we proposed a convolutional neural network called residual dense U-Net. This network is devised based on the original U-Net network. The encoder-decoder architecture in U-Net can restore the feature map to the resolution of the original image and obtain high-level semantic features. The skip-connection in U-Net can fuse the features after up-sampling and down-sampling to prevent both high-level semantic features and low-level semantic features from being lost after down-sampling. In the encoder and decoder parts, we utilize the residual dense block (RDB) from Residual Dense Network. Before each max-pooling, we replace the last convolutional layer in the original U-Net architecture with RDB. After each up-sampling, the last convolutional layer in the original U-Net architecture will also be replaced with RDB. The proposed method will be used to find poultry farms in Taiwan from satellite images. The prediction results will be evaluated using several indicators such as IOU, precision, recall, and F1-score.

Day-Yu Chao

Papers

Locating Waterfowl Farms from Satellite Images with Parallel Residual U-Net Architecture

Identifying Poultry Farms from Satellite Images with Residual Dense U-Net