DLDong Liang

Dong Liang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Dong Liang has a total of 27 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • SCULPT: Semantic-aware causal prompt tuning for out-of-distribution detection of whole slide images

    by Pengzhong Sun, Xiangyu Li, Dong Liang, Jun Liu, Zhanshi Zhu, Xiaokun Li, Suyu Dong, Gongning Luo, Wei Wang, Kuanquan Wang, et al.


  • Masked graph convolutional neural network for medical image segmentation with anatomical priors

    by Yuzhen Li, Dong Liang, Xingyu Qiu, Xiangyu Li, Kuanquan Wang, Gongning Luo


  • Open Access

    Active CT Reconstruction with a Learned Sampling Policy

    by Ce Wang, Kun Shang, Haimiao Zhang, Shang Zhao, Dong Liang, S. Kevin Zhou

    Computed tomography (CT) is a widely-used imaging technology that assists clinical decision-making with high-quality human body representations. To reduce the radiation dose posed by CT, sparse-view and limited-angle CT are developed with preserved image quality. However, these methods are still stuck with a fixed or uniform sampling strategy, which inhibits the possibility of acquiring a better image with an even reduced dose. In this paper, we explore this possibility via learning an active sampling policy that optimizes the sampling positions for patient-specific, high-quality reconstruction. To this end, we design an \textit{intelligent agent} for active recommendation of sampling positions based on on-the-fly reconstruction with obtained sinograms in a progressive fashion. With such a design, we achieve better performances on the NIH-AAPM dataset over popular uniform sampling, especially when the number of views is small. Finally, such a design also enables RoI-aware reconstruction with improved reconstruction quality within regions of interest (RoI's) that are clinically important. Experiments on the VerSe dataset demonstrate this ability of our sampling policy, which is difficult to achieve based on uniform sampling.

  • Densenet based on coordinated attention for cotton aphid disease severity classification

    by Wenxia Bao, Yuanyuan Wang, Dongyan Zhang, Hongbo Qiao, Dong Liang

    The classification of cotton aphid disease severity is conducive to the accurate control of cotton aphids. Because the cotton aphid images taken in the natural environment have some problems, such as complex background, different lighting conditions and difficult to distinguish between different grades, the existing convolutional neural networks do not have high accuracy in the classification of cotton aphid grades in the natural scene images. This study proposes an improved DenseNet classification network, CA_DenseNet_BC_100. It is used to classify the severity of cotton aphids in natural scene images. CA_DenseNet_BC_100 network inherits the advantages of DenseNet_BC_100 network design and applies a novel attention coordination module. The experimental results show that the classification accuracy of CA_DenseNet_BC_100 network for the severity of cotton aphids in natural scene images is better than that of a series of existing networks such as resnet50, ShuffleNetv2, GhostNet, MobileNetv3 and DenseNet.

  • Horizon Detection From Electro-Optical Sensors Under Maritime Environment

    by Dong Liang, Ya Liang

    In recent years, target detection in maritime surveillance has become a popular subject. Horizon under maritime environment is one of the most important semantic boundaries for segmenting the image into sea and sky. With the help of horizon, seeking scope can be restricted to a definite small area, which significantly reduces the computational complexity. Meanwhile, horizon is used to automatically adjust the attitude of air vehicles as well. Although a plenty of algorithms for horizon detection have been created, they are prone to low visibility and occlusions. In this paper, a novel robust algorithm based on probability distribution and physical characteristics is proposed. At first, the probabilities are distributed to each vertically divided region by weighted textures, and the sea-sky region is located from definite probabilistic intervals. Based on a Canny edge detector, Hough transform extracts a series of candidate horizons. Finally, a novel voting method is applied. The line with the maximum value will be regarded as the true horizon. The proposed algorithm can not only precisely detect the horizon from fine images but also from blurred image, even with splashed camera. The performance of the proposed algorithm is verified by experiments on 65 videos containing more than 20,000 frames under different sceneries. The results illustrate that our algorithm can detect the horizon line with minor errors in comparison with other five state-of-the-art algorithms.


  • Robust sea-sky-line detection for complex sea background

    by Dong Liang, Weigang Zhang, Qiaozhong Huang, Fei Yang

    Sea-sky-line detection in complicated sea-sky background plays an important role in the long-range moving objects detection and the security of sea port. This paper presents a new method to detect the sea-sky-line. In the first step, the original image is transformed into grayscale image and the sea-sky-line region is located by using the textural features gray-level co-occurrence matrix based on texture feature. In the second step, the adaptive segmentation thresholds are obtained by the OTSU algorithm, which produces a set of sea-sky-line candidate points. Finally, a simple clustering method is adopted to select appropriate points and transform them by straight line fitting. The results of experiments established on real-world sea-sky images demonstrate the effectiveness and robustness of the proposed approach.