Article
Computer Science, Artificial Intelligence
Xuebin Xu, Meng Lei, Dehua Liu, Muyu Wang, Longbin Lu
Summary: This paper presents a multi-interaction feature fusion network model based on Kiu-Net for lung segmentation in chest CT images, aiming to address the limitations of current methods using U-Net in detecting small structures and accurately segmenting boundaries. The experimental results demonstrate the superiority of this model over existing methods.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Lei Chen, Tieyong Cao, Yunfei Zheng, Zheng Fang
Summary: This paper proposes a self-distillation object segmentation method via frequency domain knowledge augmentation, which does not require complex auxiliary teacher structures and a large number of training samples. By constructing an object segmentation network that efficiently integrates multi-level features, a pixel-wise virtual teacher generation model is proposed to transfer pixel-wise knowledge to the object segmentation network through self-distillation learning, thereby improving its generalization ability. A frequency domain knowledge adaptive generation method is used to augment data, dynamically adjusting the learnable pixel-wise quantization table with a differentiable quantization operator. Experimental results show that the proposed method effectively enhances the performance of the object segmentation network, outperforming recent self-distillation methods, and achieving an average F-beta and mIoU increase of about 1.5% and 3.6% compared to a typical feature refinement self-distillation method.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Yaxin Ji, Lan Di
Summary: This paper presents a textile defect detection method that utilizes a multi-proportion spatial attention mechanism and channel memory feature fusion network. It addresses the difficulties presented by complicated shapes and large size variations, resulting in improved detection accuracy.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Shuang Cai, Shanmin Yang, Jing Hu, Xi Wu
Summary: This paper proposes a novel dual-granularity feature fusion network for VI-ReID, aiming to enhance representation and robustness by fusing global and local features. Moreover, an identity-aware modal discrepancy loss is proposed to dynamically align the cross-modal distribution of pedestrians and reduce modality discrepancies.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Chengxian Li, Ming Shao, Qirui Yang, Siyu Xia
Summary: The authors propose a novel counting model to estimate the number of repetitive actions in temporal 3D skeleton data. This is the first work of its kind using skeleton data for high-precision repetitive action counting. The model follows a bottom-up pipeline to clip the sub-action and uses robust aggregation in inference. The proposed model outperforms existing video-based methods in terms of accuracy in real-time inference.
IET COMPUTER VISION
(2023)
Review
Computer Science, Artificial Intelligence
Ren Qian, Renyan Feng, Wangduo Xie, Wenbang Yang, Yong Zhao
Summary: In the field of stereo matching, the accuracy of the disparity map depends on how well the algorithm handles ill regions. The proposed Concatenated Dilated Convolution (CDC) block and Multi-scale Feature Fusion module (MFF) effectively extract regional context by increasing the receptive field and enhancing the smoothness of the feature map. Additionally, a channel-distribution algorithm is used to reduce the number of parameters while maintaining performance. Experimental results demonstrate that MFF and CDC improve the performance of ill areas and networks with minimal parameters.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Jie Zhang, Zhichao Zhang, Hua Liu, Shiqiang Xu
Summary: Breast cancer classification and segmentation are crucial in identifying and detecting benign and malignant breast lesions. However, these tasks face challenges due to the characteristics of cancer itself and the lack of consideration of their potential relationship. To address these challenges, this paper proposes a novel Semantic-aware transformer (SaTransformer) that performs classification and segmentation simultaneously through a unified framework.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Peter Macsik, Jarmila Pavlovicova, Slavomir Kajan, Jozef Goga, Veronika Kurilova
Summary: Diabetic retinopathy (DR) is a disease that can cause irreversible eye damage and even blindness, and early diagnosis is crucial. A new deep learning approach is proposed to automatically diagnose and stage retinal images using an ensemble of different models. This study demonstrates the effectiveness of the approach in improving the prognosis of diabetic retinopathy.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Zhong Wu, Hong Zhu, Lili He, Dong Wang, Jing Shi, Wenhuan Wu
Summary: Cost aggregation is crucial for accurate stereo matching. A reasonable algorithm should aggregate costs within homogeneous regions to prevent issues like edge-fattening and loss of fine structures. To address these problems, a novel densely connected asymmetric convolution block (Dense-ACB) is proposed to explicitly construct receptive fields with various shapes, effectively mitigating issues caused by mismatching shapes of receptive fields and homogeneous regions. The proposed Dense-ACB brings new insights to CNN-based stereo matching networks.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Xinfeng Zhang, Jiaming Zhang, Yitian Zhang, Maoshen Jia, Hui Li, Xiaomin Liu
Summary: Accurate segmentation of hard exudates in early non-proliferative diabetic retinopathy can assist physicians in targeted treatment to avoid vision damage caused by disease deterioration. CT-ALUnet, an Adaptive Learning Unet-based adversarial network with Convolutional neural network and Transformer, is proposed for automatic segmentation of hard exudates.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Siyu Wang, WeiPeng Li, Ruitao Lu, Xiaogang Yang, Jianxiang Xi, Jiuan Gao
Summary: In this study, a novel neural network acceleration method based on selective activation is proposed. By using selective activation as the algebraic basis to reduce matrix multiplication calculations, and introducing an Activation Management Unit to screen and remove activated neurons and reduce the number of calculations, this method significantly reduces the number of neural network calculations while maintaining accuracy.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Chunyun Meng, Ernest Domanaanmwi Ganaa, Bin Wu, Zhen Tan, Li Luan
Summary: This paper proposes a novel framework called body topology information generation and matching (BTIGM) to address the challenge of occlusion in pedestrian image recognition. By restoring holistic pedestrian images with body topology information and utilizing cosine distance for matching, the proposed framework achieves better performance compared to existing methods. Extensive experiments on different datasets validate the effectiveness of the framework.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Cheng Yang, Guanming Lu
Summary: This study proposes an unsupervised image-blind super-resolution method called RDFL, which addresses the challenge of accurately replicating the complexity of real-world image degradation. The method utilizes degradation feature-based learning and a Transformer-based SR network to achieve high-quality image reconstruction.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Song Yan, Lei Zhang
Summary: In this paper, a novel lightweight grasping detection model is proposed, which addresses the theoretical challenges associated with grasping detection by incorporating attention mechanisms and residual modules. The model achieves remarkable detection accuracy rates on the Cornell Grasp dataset, demonstrating its exceptional performance in overcoming the theoretical complexities of grasping detection.
IET COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Yitao Ren, Peiyang Jin, Yiyang Li, Keming Mao
Summary: This paper proposes an effective ghost module based spectral network for hyperspectral image classification. It adopts Ghost3D module to reduce model parameter size by generating redundant feature maps with linear transformation. Ghost2D module with channel-wise attention is used to explore informative spectral feature representation. Compared with existing methods, the proposed approach achieves superior performance on three hyperspectral image datasets with fewer sample labelling and less resource consumption.
IET IMAGE PROCESSING
(2023)