2025
AdaTok: Adaptive Token Compression with Object-Aware Representations for Efficient Multimodal LLMs
Xinliang Zhang, Lei Zhu, Hangzhou He, Shuang Zeng, Ourui Fu, Jiakui Hu, Zhengjian Yao, Yanye Lu
Submitted to CVPR'26 2025
In this study, we propose a object-level visual representation compression strategy for multimodal large language models.
AdaTok: Adaptive Token Compression with Object-Aware Representations for Efficient Multimodal LLMs
Xinliang Zhang, Lei Zhu, Hangzhou He, Shuang Zeng, Ourui Fu, Jiakui Hu, Zhengjian Yao, Yanye Lu
Submitted to CVPR'26 2025
In this study, we propose a object-level visual representation compression strategy for multimodal large language models.
Branches Mutual Promotion for End-to-End Weakly Supervised Semantic Segmentation
Kaiwen Li, Hangzhou He, Shuang Zeng, Xinliang Zhang, Yuanwei Li, Lei Zhu, Yanye Lu
IEEE Transactions on Medical Imaging(TMI) 2025 中科院一区Top
In this paper, we introduce point annotations to fundus vessel segmentation and propose a novel method, called Points-based Vessel segmentation Network (PVN), to enhance the segmentation accuracy.
Branches Mutual Promotion for End-to-End Weakly Supervised Semantic Segmentation
Kaiwen Li, Hangzhou He, Shuang Zeng, Xinliang Zhang, Yuanwei Li, Lei Zhu, Yanye Lu
IEEE Transactions on Medical Imaging(TMI) 2025 中科院一区Top
In this paper, we introduce point annotations to fundus vessel segmentation and propose a novel method, called Points-based Vessel segmentation Network (PVN), to enhance the segmentation accuracy.
Novel Extraction of Discriminative Fine-Grained Feature to Improve Retinal Vessel Segmentation
Kaiwen Li, Hangzhou He, Shuang Zeng, Xinliang Zhang, Yuanwei Li, Lei Zhu, Yanye Lu
IEEE Transactions on Medical Imaging(TMI) 2025
In this paper, we propose a novel Attention U-shaped Kolmogorov–Arnold Network named AttUKAN along with a novel Label-guided Pixel-wise Contrastive Loss for retinal vessel segmentation.
Novel Extraction of Discriminative Fine-Grained Feature to Improve Retinal Vessel Segmentation
Kaiwen Li, Hangzhou He, Shuang Zeng, Xinliang Zhang, Yuanwei Li, Lei Zhu, Yanye Lu
IEEE Transactions on Medical Imaging(TMI) 2025
In this paper, we propose a novel Attention U-shaped Kolmogorov–Arnold Network named AttUKAN along with a novel Label-guided Pixel-wise Contrastive Loss for retinal vessel segmentation.
Branches Mutual Promotion for End-to-End Weakly Supervised Semantic Segmentation
Lei Zhu, Xinliang Zhang, Hangzhou He, Qian Chen, Sha Li, Shuang Zeng, Yibao Zhang, Qiushi Ren, Yanye Lu
IEEE Transactions on Neural Networks and Learning Systems(TNNLS) 2025 中科院一区Top
Previous methods online-trained classification branch to provide pseudo annotations for supervising the segmentation branch, which makes the classification branch dominate the whole concurrent training process. We propose a bidirectional supervision mechanism to achieve mutual promotion for End2End weakly supervised semantic segmentation field.
Branches Mutual Promotion for End-to-End Weakly Supervised Semantic Segmentation
Lei Zhu, Xinliang Zhang, Hangzhou He, Qian Chen, Sha Li, Shuang Zeng, Yibao Zhang, Qiushi Ren, Yanye Lu
IEEE Transactions on Neural Networks and Learning Systems(TNNLS) 2025 中科院一区Top
Previous methods online-trained classification branch to provide pseudo annotations for supervising the segmentation branch, which makes the classification branch dominate the whole concurrent training process. We propose a bidirectional supervision mechanism to achieve mutual promotion for End2End weakly supervised semantic segmentation field.
V2C-CBM: Building Concept Bottlenecks with Vision-to-Concept Tokenizer
Hangzhou He, Lei Zhu, Xinliang Zhang, Shuang Zeng, Qian Chen, Yanye Lu
Association for the Advancement of Artificial Intelligence (AAAI) 2025 Imageomics Oral
we adopt common words as base concept vocabulary and leverage auxiliary unlabeled images to construct a Vision-to-Concept (V2C) tokenizer that can explicitly quantize images into their most relevant visual concepts, thus creating a vision-oriented concept bottleneck tightly coupled with the multimodal model.
V2C-CBM: Building Concept Bottlenecks with Vision-to-Concept Tokenizer
Hangzhou He, Lei Zhu, Xinliang Zhang, Shuang Zeng, Qian Chen, Yanye Lu
Association for the Advancement of Artificial Intelligence (AAAI) 2025 Imageomics Oral
we adopt common words as base concept vocabulary and leverage auxiliary unlabeled images to construct a Vision-to-Concept (V2C) tokenizer that can explicitly quantize images into their most relevant visual concepts, thus creating a vision-oriented concept bottleneck tightly coupled with the multimodal model.
Generative learning-based lightweight MRI brain tumor segmentation with missing modalities
Xinliang Zhang, Qian Chen, Hangzhou He, Lei Zhu, Zhaoheng Xie, Yanye Lu
Expert Systems with Applications 2025 中科院一区Top
In this study, we propose a class-driven scribble promotion network, which utilizes both scribble annotations and pseudo-labels informed by image-level classes and global semantics for supervision.
Generative learning-based lightweight MRI brain tumor segmentation with missing modalities
Xinliang Zhang, Qian Chen, Hangzhou He, Lei Zhu, Zhaoheng Xie, Yanye Lu
Expert Systems with Applications 2025 中科院一区Top
In this study, we propose a class-driven scribble promotion network, which utilizes both scribble annotations and pseudo-labels informed by image-level classes and global semantics for supervision.
2024
Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label
Xinliang Zhang, Lei Zhu, Hangzhou He, Lujia Jin, Yanye Lu
Association for the Advancement of Artificial Intelligence (AAAI) 2024 CCF-A
In this study, we propose a class-driven scribble promotion network, which utilizes both scribble annotations and pseudo-labels informed by image-level classes and global semantics for supervision.
Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label
Xinliang Zhang, Lei Zhu, Hangzhou He, Lujia Jin, Yanye Lu
Association for the Advancement of Artificial Intelligence (AAAI) 2024 CCF-A
In this study, we propose a class-driven scribble promotion network, which utilizes both scribble annotations and pseudo-labels informed by image-level classes and global semantics for supervision.