Yi Li
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
Yi Li has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
- Licheng Jiao
Xidian University, Xian, China
6 shared publications - Shuang Wang
School of Artificial Intelligence, Xidian University, Xi’an 710126, China
6 shared publications - Dou Quan
Key Laboratory of Intelligent Perception and Image Understanding of Ministry of Education, School of Artificial Intelligence, Xidian University, Xian, China
5 shared publications - Jocelyn Chanussot
Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK
4 shared publications - Yabo Yan
Department of Orthopedics, Xijing Hospital, Air Force Military Medical University, Xi’an, China
4 shared publications - Biao Hou
School of Artificial Intelligence, Xidian University, Xi’an 710071, China
2 shared publications
Latest Publications
- 11 July 2025
- Published by Springer Nature in European Spine Journal
- p. 1-11
A novel artificial Intelligence-Based model for automated Lenke classification in adolescent idiopathic scoliosis
Purpose To develop an artificial intelligence (AI)-driven model for automatic Lenke classification of adolescent idiopathic scoliosis (AIS) and assess its performance. Methods This retrospective study utilized 860 spinal radiographs from 215 AIS patients with four views, including 161 training sets and 54 testing sets. Additionally, 1220 spinal radiographs from 610 patients with only anterior-posterior (AP) and lateral (LAT) views were collected for training. The model was designed to perform keypoint detection, pedicle segmentation, and AIS classification based on a custom classification strategy. Its performance was evaluated against the gold standard using metrics such as mean absolute difference (MAD), intraclass correlation coefficient (ICC), Bland-Altman plots, Cohen’s Kappa, and the confusion matrix. Results In comparison to the gold standard, the MAD for all predicted angles was 2.29°, with an excellent ICC. Bland-Altman analysis revealed minimal differences between the methods. For Lenke classification, the model exhibited exceptional consistency in curve type, lumbar modifier, and thoracic sagittal profile, with average Kappa values of 0.866, 0.845, and 0.827, respectively, and corresponding accuracy rates of 87.07%, 92.59%, and 92.59%. Subgroup analysis further confirmed the model’s high consistency, with Kappa values ranging from 0.635 to 0.930, 0.672 to 0.926, and 0.815 to 0.847, and accuracy rates between 90.7 and 98.1%, 92.6–98.3%, and 92.6–98.1%, respectively. Conclusion This novel AI system facilitates the rapid and accurate automatic Lenke classification, offering potential assistance to spinal surgeons.
- 11 July 2025
- Published by Springer Nature in European Spine Journal
- p. 1-11
A novel artificial Intelligence-Based model for automated Lenke classification in adolescent idiopathic scoliosis
Purpose To develop an artificial intelligence (AI)-driven model for automatic Lenke classification of adolescent idiopathic scoliosis (AIS) and assess its performance. Methods This retrospective study utilized 860 spinal radiographs from 215 AIS patients with four views, including 161 training sets and 54 testing sets. Additionally, 1220 spinal radiographs from 610 patients with only anterior-posterior (AP) and lateral (LAT) views were collected for training. The model was designed to perform keypoint detection, pedicle segmentation, and AIS classification based on a custom classification strategy. Its performance was evaluated against the gold standard using metrics such as mean absolute difference (MAD), intraclass correlation coefficient (ICC), Bland-Altman plots, Cohen’s Kappa, and the confusion matrix. Results In comparison to the gold standard, the MAD for all predicted angles was 2.29°, with an excellent ICC. Bland-Altman analysis revealed minimal differences between the methods. For Lenke classification, the model exhibited exceptional consistency in curve type, lumbar modifier, and thoracic sagittal profile, with average Kappa values of 0.866, 0.845, and 0.827, respectively, and corresponding accuracy rates of 87.07%, 92.59%, and 92.59%. Subgroup analysis further confirmed the model’s high consistency, with Kappa values ranging from 0.635 to 0.930, 0.672 to 0.926, and 0.815 to 0.847, and accuracy rates between 90.7 and 98.1%, 92.6–98.3%, and 92.6–98.1%, respectively. Conclusion This novel AI system facilitates the rapid and accurate automatic Lenke classification, offering potential assistance to spinal surgeons.
- Open Access
- 3 February 2025
- Published by Elsevier in Neurocomputing
- Vol. 626
FASTCC: A lightweight human pose detection method leveraging SimCC
- 3 February 2025
- Published by Elsevier in Neurocomputing
- Vol. 626
Open AccessFASTCC: A lightweight human pose detection method leveraging SimCC
- 2 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-13
LM-Net: A Lightweight Matching Network for Remote Sensing Image Matching and Registration
Deep feature learning methods have shown significant advantages over handcrafted feature-based methods in remote sensing image matching and registration. Existing deep learning methods usually introduce complex modules into the deep convolutional network for more robust feature learning. However, they usually require high computation and memory resources for the computing device and have expensive time costs for image registration. As a basic image processing task, it is crucial to build a lightweight matching network for fast and accurate image matching and registration. Unfortunately, the image matching performance will decrease significantly when we directly compress the deep model to a lightweight one. This paper proposes a lightweight matching network (LM-Net) based on the knowledge distillation (KD) learning framework for remote sensing image matching and registration. We first build an LM-Net with three convolutional layers. Then, this paper proposes an effective knowledge distillation approach for network optimization, which transfers the effective knowledge from the deep matching network to LM-Net to improve image matching performances. Specifically, this paper considers the useful information in the instance samples and the relation information between samples. It designs the feature and feature relation distillation learning for LM-Net training. Extensive experimental results and analysis have shown the effectiveness and advantages of the proposed LM-Net. LM-Net can reduce the number of parameters and computational complexity of the matching network. Meanwhile, LM-Net can significantly decrease the time cost and achieve results comparable to those of the deep model. It reduces the average image registration time by 42% on remote sensing image matching and registration. Additionally, LM-Net generalizes well on other multi-modal remote sensing images.
- 2 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-13
LM-Net: A Lightweight Matching Network for Remote Sensing Image Matching and Registration
Deep feature learning methods have shown significant advantages over handcrafted feature-based methods in remote sensing image matching and registration. Existing deep learning methods usually introduce complex modules into the deep convolutional network for more robust feature learning. However, they usually require high computation and memory resources for the computing device and have expensive time costs for image registration. As a basic image processing task, it is crucial to build a lightweight matching network for fast and accurate image matching and registration. Unfortunately, the image matching performance will decrease significantly when we directly compress the deep model to a lightweight one. This paper proposes a lightweight matching network (LM-Net) based on the knowledge distillation (KD) learning framework for remote sensing image matching and registration. We first build an LM-Net with three convolutional layers. Then, this paper proposes an effective knowledge distillation approach for network optimization, which transfers the effective knowledge from the deep matching network to LM-Net to improve image matching performances. Specifically, this paper considers the useful information in the instance samples and the relation information between samples. It designs the feature and feature relation distillation learning for LM-Net training. Extensive experimental results and analysis have shown the effectiveness and advantages of the proposed LM-Net. LM-Net can reduce the number of parameters and computational complexity of the matching network. Meanwhile, LM-Net can significantly decrease the time cost and achieve results comparable to those of the deep model. It reduces the average image registration time by 42% on remote sensing image matching and registration. Additionally, LM-Net generalizes well on other multi-modal remote sensing images.
- 19 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-16
A Unified Deep Learning Network for Remote Sensing Image Registration and Change Detection
Image registration and change detection are crucial for multi-temporal remote sensing image analysis. The images should be registered before the change information detection. Existing deep learning methods have shown significant advantages in image registration and change detection tasks. They usually design two independent task-specific deep networks for image registration and change detection, respectively. These independent deep networks will learn from scratch and rely on many task-specific labeled training datasets. This paper finds that image registration and change detection have similar learning mechanisms, which focus on extracting discriminative features. Inspired by this, we propose a unified deep learning network, URCNet, that can perform image alignment and change information detection through a single network. Additionally, this paper proposes various deep collaborative learning methods for URCNet optimization, which enforce that the URCNet can effectively support remote sensing image registration and change detection simultaneously. Extensive experiments demonstrate the effectiveness of the proposed URCNet for image registration and change detection, which can achieve comparable and better results with task-specific and more complex deep networks. The proposed URCNet can support multi-tasks based on the same scene images, different scene images, and even multi-modal images. Moreover, URCNet shows significant advantages over other deep networks in change detection under limited labeled datasets.
- 19 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-16
A Unified Deep Learning Network for Remote Sensing Image Registration and Change Detection
Image registration and change detection are crucial for multi-temporal remote sensing image analysis. The images should be registered before the change information detection. Existing deep learning methods have shown significant advantages in image registration and change detection tasks. They usually design two independent task-specific deep networks for image registration and change detection, respectively. These independent deep networks will learn from scratch and rely on many task-specific labeled training datasets. This paper finds that image registration and change detection have similar learning mechanisms, which focus on extracting discriminative features. Inspired by this, we propose a unified deep learning network, URCNet, that can perform image alignment and change information detection through a single network. Additionally, this paper proposes various deep collaborative learning methods for URCNet optimization, which enforce that the URCNet can effectively support remote sensing image registration and change detection simultaneously. Extensive experiments demonstrate the effectiveness of the proposed URCNet for image registration and change detection, which can achieve comparable and better results with task-specific and more complex deep networks. The proposed URCNet can support multi-tasks based on the same scene images, different scene images, and even multi-modal images. Moreover, URCNet shows significant advantages over other deep networks in change detection under limited labeled datasets.
- 24 August 2023
- Published by Springer Nature in European Journal of Pediatrics
- Vol. 182 (11) , 4983-4991
A novel approach for screening standard anteroposterior pelvic radiographs in children
Anteroposterior pelvic radiography is the first‐line imaging modality for diagnosing developmental dysplasia of the hip (DDH). Nonstandard radiographs with pelvic malposition make the correct diagnosis of DDH challenging. However, as the only method available for screening standard pelvic radiographs, traditional manual assessment is relatively laborious and potentially erroneous. We retrospectively collected 3,247 pelvic radiographs. There were 2,887 radiographs randomly selected to train and optimize the AI model. Then 362 radiographs were used to test the model’s diagnostic performance. Its diagnostic accuracy was assessed using receiver operating characteristic (ROC) curves and measurement consistency using Bland–Altman plots. In 362 radiographs, the AI model’s area under ROC curves, accuracy, sensitivity, and specificity for quality assessment was 0.993, 99.4% (360/362), 98.6% (138/140), and 100.0% (222/222), respectively. Compared with clinicians, the 95% limits of agreement (Bland–Altman analysis) for pelvic tilt index (PTI) and pelvic rotation index (PRI), as determined by the model, were -0.052–0.072 and -0.088–0.055, respectively. Conclusions: The artificial intelligence-assisted method was more efficient and highly consistent with clinical experts. This method can be used for real-time validation of the quality of pelvic radiographs in current picture archiving and communications systems (PACS). What is Known: • Nonstandard pediatric radiographs with pelvic malposition make the correct diagnosis of developmental dysplasia of the hip (DDH) challenging. • Traditional manual assessment remains the only method available for screening standard pediatric pelvic radiographs, which is relatively laborious and potentially erroneous. What is New: • This study proposed an artificial intelligence-assisted model to assess the quality of pediatric pelvic radiographs accurately and efficiently. • We recommend the integration of the model into current picture archiving and communications systems (PACS) for real-time screening of standard pediatric pelvic radiographs.
- 24 August 2023
- Published by Springer Nature in European Journal of Pediatrics
- Vol. 182 (11) , 4983-4991
A novel approach for screening standard anteroposterior pelvic radiographs in children
Anteroposterior pelvic radiography is the first‐line imaging modality for diagnosing developmental dysplasia of the hip (DDH). Nonstandard radiographs with pelvic malposition make the correct diagnosis of DDH challenging. However, as the only method available for screening standard pelvic radiographs, traditional manual assessment is relatively laborious and potentially erroneous. We retrospectively collected 3,247 pelvic radiographs. There were 2,887 radiographs randomly selected to train and optimize the AI model. Then 362 radiographs were used to test the model’s diagnostic performance. Its diagnostic accuracy was assessed using receiver operating characteristic (ROC) curves and measurement consistency using Bland–Altman plots. In 362 radiographs, the AI model’s area under ROC curves, accuracy, sensitivity, and specificity for quality assessment was 0.993, 99.4% (360/362), 98.6% (138/140), and 100.0% (222/222), respectively. Compared with clinicians, the 95% limits of agreement (Bland–Altman analysis) for pelvic tilt index (PTI) and pelvic rotation index (PRI), as determined by the model, were -0.052–0.072 and -0.088–0.055, respectively. Conclusions: The artificial intelligence-assisted method was more efficient and highly consistent with clinical experts. This method can be used for real-time validation of the quality of pelvic radiographs in current picture archiving and communications systems (PACS). What is Known: • Nonstandard pediatric radiographs with pelvic malposition make the correct diagnosis of developmental dysplasia of the hip (DDH) challenging. • Traditional manual assessment remains the only method available for screening standard pediatric pelvic radiographs, which is relatively laborious and potentially erroneous. What is New: • This study proposed an artificial intelligence-assisted model to assess the quality of pediatric pelvic radiographs accurately and efficiently. • We recommend the integration of the model into current picture archiving and communications systems (PACS) for real-time screening of standard pediatric pelvic radiographs.
- 26 May 2023
- Published by SPIE-Intl Soc Optical Eng
- Vol. 12700, 538-542
Using a region-restrictive erasing method for one-stage weakly-supervised semantic segmentation
Image segmentation is a classical and basic problem in the domain of computer vision. Due to the fact that fully supervision segmentation methods require dense time-consuming and expensive manual-annotations, lots of WSSS (Weakly-Supervised Semantic Segmentation) methods have been proposed to take advantage of the simplicity and availability of weak supervision annotations. In this work, we build an integrated framework to jointly train the classification task and the segmentation task guided by a self-supervised thinking with only image-level supervision and a compound refinement strategy. Then, we introduce a restrictive adversarial erasing approach to push our model to find more segmentation cues. We evaluate the proposed method on PASCAL VOC 2012 benchmark, and the experiments show that our method can achieve competitive performance compared with the earlier methods.
- 26 May 2023
- Published by SPIE-Intl Soc Optical Eng
- Vol. 12700, 538-542
Using a region-restrictive erasing method for one-stage weakly-supervised semantic segmentation
Image segmentation is a classical and basic problem in the domain of computer vision. Due to the fact that fully supervision segmentation methods require dense time-consuming and expensive manual-annotations, lots of WSSS (Weakly-Supervised Semantic Segmentation) methods have been proposed to take advantage of the simplicity and availability of weak supervision annotations. In this work, we build an integrated framework to jointly train the classification task and the segmentation task guided by a self-supervised thinking with only image-level supervision and a compound refinement strategy. Then, we introduce a restrictive adversarial erasing approach to push our model to find more segmentation cues. We evaluate the proposed method on PASCAL VOC 2012 benchmark, and the experiments show that our method can achieve competitive performance compared with the earlier methods.
- Open Access
- 16 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 16 (19391404) , 4739-4754
Efficient and Robust: A Cross-Modal Registration Deep Wavelet Learning Method for Remote Sensing Images
Deep convolutional networks are powerful for local feature learning and have shown advantages in image matching and registration. However, the significant differences between cross-modal images increase the challenge of image registration. The deep network should extract modality-invariant features to identify the matching samples and discriminative features to separate the non-matching samples. The deep network can extract features invariant to the image modality changes by multiple non-linear mapping layers. However, it does not inevitably lose rich details and affect the discrimination of features, degrading registration performances. This paper proposes a novel deep wavelet learning network (DW-Net) for local feature learning. It incorporates spectral information into deep convolutional features for improving cross-modal image matching and registration. Specifically, this paper aims to learn the multi-resolution wavelet features through multi-level wavelet transform (WT) and the convolutional network. The cross-modal images are divided into low-frequency and high-frequency parts through WT. DW-Net can adaptively extract the shared features from the low-frequency part and useful details from the high-frequency part, which can enhance the modality-invariance and discrimination of features. Additionally, the multi-resolution wavelet features contain multi-scale information and contribute to improving the matching accuracy. Extensive experiments demonstrate the significant advantages in terms of the accuracy and robustness of DW-Net on cross-modal remote sensing image registration. DW-Net can increase the image patch matching accuracy by 3.7% and improve image registration probability by 12.1%. Moreover, DW-Net shows strong generalization performances from low-resolution to high-resolution and from optical-SAR to other cross-modal image registration.
- 16 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 16 (19391404) , 4739-4754
Open AccessEfficient and Robust: A Cross-Modal Registration Deep Wavelet Learning Method for Remote Sensing Images
Deep convolutional networks are powerful for local feature learning and have shown advantages in image matching and registration. However, the significant differences between cross-modal images increase the challenge of image registration. The deep network should extract modality-invariant features to identify the matching samples and discriminative features to separate the non-matching samples. The deep network can extract features invariant to the image modality changes by multiple non-linear mapping layers. However, it does not inevitably lose rich details and affect the discrimination of features, degrading registration performances. This paper proposes a novel deep wavelet learning network (DW-Net) for local feature learning. It incorporates spectral information into deep convolutional features for improving cross-modal image matching and registration. Specifically, this paper aims to learn the multi-resolution wavelet features through multi-level wavelet transform (WT) and the convolutional network. The cross-modal images are divided into low-frequency and high-frequency parts through WT. DW-Net can adaptively extract the shared features from the low-frequency part and useful details from the high-frequency part, which can enhance the modality-invariance and discrimination of features. Additionally, the multi-resolution wavelet features contain multi-scale information and contribute to improving the matching accuracy. Extensive experiments demonstrate the significant advantages in terms of the accuracy and robustness of DW-Net on cross-modal remote sensing image registration. DW-Net can increase the image patch matching accuracy by 3.7% and improve image registration probability by 12.1%. Moreover, DW-Net shows strong generalization performances from low-resolution to high-resolution and from optical-SAR to other cross-modal image registration.
- Open Access
- 30 March 2023
- Published by Frontiers Media SA in Frontiers in Pediatrics
- Vol. 11, 1080194
Clinical thought-based software for diagnosing developmental dysplasia of the hip on pediatric pelvic radiographs
Background: The common methods of radiographic diagnosis of developmental dysplasia of the hip (DDH) include measuring hip parameters and quantifying the degree of hip dislocation. However, clinical thought-based analysis of hip parameters may be a more effective way to achieve expert-like diagnoses of DDH. This study aims to develop a diagnostic strategy-based software for pediatric DDH and validate its clinical feasibility.Methods: In total, 543 anteroposterior pelvic radiographs were retrospectively collected from January 2017 to December 2021. Two independent clinicians measured four diagnostic indices to compare the diagnoses made by the software and conventional manual method. The diagnostic accuracy was evaluated using the receiver operator characteristic (ROC) curves and confusion matrix, and the consistency of parametric measurements was assessed using Bland-Altman plots.Results: In 543 cases (1,086 hips), the area under the curve, accuracy, sensitivity, and specificity of the software for diagnosing DDH were 0.988–0.994, 99.08%–99.72%, 98.07%–100.00%, and 99.59%, respectively. Compared with the expert panel, the Bland-Altman 95% limits of agreement for the acetabular index, as determined by the software, were −2.09°–2.91° (junior orthopedist) and −1.98°–2.72° (intermediate orthopedist). As for the lateral center-edge angle, the 95% limits were −3.68°–5.28° (junior orthopedist) and −2.94°–4.59° (intermediate orthopedist).Conclusions: The software can provide expert-like analysis of pelvic radiographs and obtain the radiographic diagnosis of pediatric DDH with great consistency and efficiency. Its initial success lays the groundwork for developing a full-intelligent comprehensive diagnostic system of DDH.
- 30 March 2023
- Published by Frontiers Media SA in Frontiers in Pediatrics
- Vol. 11, 1080194
Open AccessClinical thought-based software for diagnosing developmental dysplasia of the hip on pediatric pelvic radiographs
Background: The common methods of radiographic diagnosis of developmental dysplasia of the hip (DDH) include measuring hip parameters and quantifying the degree of hip dislocation. However, clinical thought-based analysis of hip parameters may be a more effective way to achieve expert-like diagnoses of DDH. This study aims to develop a diagnostic strategy-based software for pediatric DDH and validate its clinical feasibility.Methods: In total, 543 anteroposterior pelvic radiographs were retrospectively collected from January 2017 to December 2021. Two independent clinicians measured four diagnostic indices to compare the diagnoses made by the software and conventional manual method. The diagnostic accuracy was evaluated using the receiver operator characteristic (ROC) curves and confusion matrix, and the consistency of parametric measurements was assessed using Bland-Altman plots.Results: In 543 cases (1,086 hips), the area under the curve, accuracy, sensitivity, and specificity of the software for diagnosing DDH were 0.988–0.994, 99.08%–99.72%, 98.07%–100.00%, and 99.59%, respectively. Compared with the expert panel, the Bland-Altman 95% limits of agreement for the acetabular index, as determined by the software, were −2.09°–2.91° (junior orthopedist) and −1.98°–2.72° (intermediate orthopedist). As for the lateral center-edge angle, the 95% limits were −3.68°–5.28° (junior orthopedist) and −2.94°–4.59° (intermediate orthopedist).Conclusions: The software can provide expert-like analysis of pelvic radiographs and obtain the radiographic diagnosis of pediatric DDH with great consistency and efficiency. Its initial success lays the groundwork for developing a full-intelligent comprehensive diagnostic system of DDH.
- Open Access
- 23 November 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 31 (10577149) , 7403-7418
Cluster Alignment With Target Knowledge Mining for Unsupervised Domain Adaptation Semantic Segmentation
Unsupervised domain adaptation (UDA) carries out knowledge transfer from the labeled source domain to the unlabeled target domain. Existing feature alignment methods in UDA semantic segmentation achieve this goal by aligning the feature distribution between domains. However, these feature alignment methods ignore the domain-specific knowledge of the target domain. In consequence, 1) the correlation among pixels of the target domain is not explored; and 2) the classifier is not explicitly designed for the target domain distribution. To conquer these obstacles, we propose a novel cluster alignment framework, which mines the domain-specific knowledge when performing the alignment. Specifically, we design a multi-prototype clustering strategy to make the pixel features within the same class tightly distributed for the target domain. Subsequently, a contrastive strategy is developed to align the distributions between domains, with the clustered structure maintained. After that, a novel affinity-based normalized cut loss is devised to learn task-specific decision boundaries. Our method enhances the model’s adaptability in the target domain, and can be used as a pre-adaptation for self-training to boost its performance. Sufficient experiments prove the effectiveness of our method against existing state-of-the-art methods on representative UDA benchmarks.
- 23 November 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 31 (10577149) , 7403-7418
Open AccessCluster Alignment With Target Knowledge Mining for Unsupervised Domain Adaptation Semantic Segmentation
Unsupervised domain adaptation (UDA) carries out knowledge transfer from the labeled source domain to the unlabeled target domain. Existing feature alignment methods in UDA semantic segmentation achieve this goal by aligning the feature distribution between domains. However, these feature alignment methods ignore the domain-specific knowledge of the target domain. In consequence, 1) the correlation among pixels of the target domain is not explored; and 2) the classifier is not explicitly designed for the target domain distribution. To conquer these obstacles, we propose a novel cluster alignment framework, which mines the domain-specific knowledge when performing the alignment. Specifically, we design a multi-prototype clustering strategy to make the pixel features within the same class tightly distributed for the target domain. Subsequently, a contrastive strategy is developed to align the distributions between domains, with the clustered structure maintained. After that, a novel affinity-based normalized cut loss is devised to learn task-specific decision boundaries. Our method enhances the model’s adaptability in the target domain, and can be used as a pre-adaptation for self-training to boost its performance. Sufficient experiments prove the effectiveness of our method against existing state-of-the-art methods on representative UDA benchmarks.
- 8 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35 (3) , 3560-3574
A Concurrent Multiscale Detector for End-to-End Image Matching
This article focuses on end-to-end image matching through joint key-point detection and descriptor extraction. To find repeatable and high discrimination key points, we improve the deep matching network from the perspectives of network structure and network optimization. First, we propose a concurrent multiscale detector (CS-det) network, which consists of several parallel convolutional networks to extract multiscale features and multilevel discriminative information for key-point detection. Moreover, we introduce an attention module to fuse the response maps of various features adaptively. Importantly, we propose two novel rank consistent losses (RC-losses) for network optimization, significantly improving image matching performances. On the one hand, we propose a score rank consistent loss (RC-S-loss) to ensure that the key points have high repeatability. Different from the score difference loss merely focusing on the absolute score of an individual key point, our proposed RC-S-loss pays more attention to the relative score of key points in the image. On the other hand, we propose a score-discrimination RC-loss to ensure that the key point has high discrimination, which can reduce the confusion from other key points in subsequent matching and then further enhance the accuracy of image matching. Extensive experimental results demonstrate that the proposed CS-det improves the mean matching result of deep detector by 1.4%–2.1%, and the proposed RC-losses can boost the matching performances by 2.7%–3.4% than score difference loss. Our source codes are available at https://github.com/iquandou/CS-Net.
- 8 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35 (3) , 3560-3574
A Concurrent Multiscale Detector for End-to-End Image Matching
This article focuses on end-to-end image matching through joint key-point detection and descriptor extraction. To find repeatable and high discrimination key points, we improve the deep matching network from the perspectives of network structure and network optimization. First, we propose a concurrent multiscale detector (CS-det) network, which consists of several parallel convolutional networks to extract multiscale features and multilevel discriminative information for key-point detection. Moreover, we introduce an attention module to fuse the response maps of various features adaptively. Importantly, we propose two novel rank consistent losses (RC-losses) for network optimization, significantly improving image matching performances. On the one hand, we propose a score rank consistent loss (RC-S-loss) to ensure that the key points have high repeatability. Different from the score difference loss merely focusing on the absolute score of an individual key point, our proposed RC-S-loss pays more attention to the relative score of key points in the image. On the other hand, we propose a score-discrimination RC-loss to ensure that the key point has high discrimination, which can reduce the confusion from other key points in subsequent matching and then further enhance the accuracy of image matching. Extensive experimental results demonstrate that the proposed CS-det improves the mean matching result of deep detector by 1.4%–2.1%, and the proposed RC-losses can boost the matching performances by 2.7%–3.4% than score difference loss. Our source codes are available at https://github.com/iquandou/CS-Net.