Chengzhi Lyu

Mr. Chengzhi Lyu

artificial intelligence
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
24 February 2020 in IEEE Access
Verified
Journal Article
Open access
CLChengzhi Lyu
Chengzhi Lyu
GHGuoqing Hu
Guoqing Hu
DWDan Wang
Dan Wang
Accurate segmentation of fine-grained information is an important step in medical image analysis applications. With the development of the encoder-decoder-based networks, various network structures and algorithms have made significant progress in semantic segmentation tasks. This work aims to present a novel high-resolution encoder-decoder network (HRED-Net) for fine-grained image segmentation that is highly accurate for small-scale targets. We design a multiscale context connection module to extract feature information without reducing the resolution, and propose a multiresolution fusion model to fine-tune the final results. In addition, these modules are trained together with a detail-oriented loss function to enhance the model's perception of fine-grained parts. Through experiments on the DRIVE dataset, we found a balance between these modules, and our comparison results show that in addition to the extraction multiscale features, the fusion of multiresolution prediction information is also beneficial for fine-grained segmentation. Our method yielded significant improvements in the accuracy and sensitivity in retinal vessel and lung segmentation tasks.

21 December 2020 in The Visual Computer
Verified
Journal Article
Closed access
PDF file
CLChengzhi Lyu
Chengzhi Lyu
GHGuoqing Hu
Guoqing Hu
DWDan Wang
Dan Wang

27 March 2024 in The Visual Computer
Verified
Journal Article
Closed access
PDF file
SLShaokang Li
Shaokang Li
CLChengzhi Lyu
Chengzhi Lyu
BXBin Xia
Bin Xia
ZCZiheng Chen
Ziheng Chen
LZLei Zhang
Lei Zhang

1 June 2020
Verified
Conference Paper
Closed access
CLChengzhi Lyu
Chengzhi Lyu
GHGuoqing Hu
Guoqing Hu
DWDan Wang
Dan Wang
Training deeper networks make an active research topic, but the appropriate is the best for specific tasks. In this paper, we proposed to design a good and stable convolutional neural network to solve many image segmentation tasks. We designed an optional skip connection module, which contains three convolutional holes. We use different sizes of convolution kernels to obtain different size of views. This module connects up sampling and down sampling layers of the same size, and dynamically adjust the parameters of the model through the skip gate unit on each convolution kernel. Our experiments on retinal and lung segmentation tasks to confirm the usability of the module, and the model automatically selects the appropriate convolutional layer and achieves a specific segmentation effect after training. The experiments prove the feasibility of the scheme and provide help for further optimization of the segmentation model with a suitable convolutional layers and parameters.

1 June 2020
Verified
Conference Paper
Closed access
DWDan Wang
Dan Wang
GHGuoqing Hu
Guoqing Hu
CLChengzhi Lyu
Chengzhi Lyu
Interactive image segmentation has gained much popularity since it can enhance the segmentation performance with human interactions. In this paper, we propose a novel framework that can effectively estimate the binary labels based on Conditional Random Field (CRF) models combining region-level information with pixels-level information. The user scribbles are treated as soft constraints. We use pixel-layer with local CRF and region-layer with fully connected CRF to model the graph. Then we combine the pixel information and region information together to calculate the pixels' likelihood term. In contrast to using relations between the pixel-layer and the region-layer during the probability computation, we calculate the probabilities of pixels and regions separately to reduce computational complexity. After the likelihood computation, a graph cut based model is used to refine the pixel-level label predictions of the image in order to avoid label inconsistency in regions and capture the boundary details. Due to the less relationship between layers and utilizing the refinement, the proposed model can provide high accuracy and user inputs insensitive results. Our experiments show that the proposed approach can efficiently and accurately segment the images compared with other state of art methods with only a few scribbles. Graph cut based fine tuning improves our segmentation accuracy and our method needs less computational cost than traditional multi-layer-based methods.

23 June 2020 in IEEE Access
Verified
Journal Article
Open access
DWDan Wang
Dan Wang
GHGuoqing Hu
Guoqing Hu
CLChengzhi Lyu
Chengzhi Lyu
Due to the irregularity and inconsistency of 3D point clouds, it is difficult to extract features directly from them. Existing methods usually extract point features independently and then use the max-pooling operation to aggregate local features, which limits the feature representation capability of their models. In this work, we design a novel spatial-related correlation path, which considers both spatial information and point correlations, to preserve high dimensional features, thereby capturing fine-detail information and long-distance context of the point cloud. We further propose a new network to aggregate the spatial aware correlations with point-wise features and global features in a learnable way. The experimental results show that our method can achieve better performance than the state-of-the-art approaches on challenging datasets. We can achieve 0.934 accuracy on ModelNet40 dataset and 0.875 mean IoU (Intersection over Union) on ShapeNet dataset with only about 2.42 million parameters.