CCCunjian Chen

Cunjian Chen

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Cunjian Chen has a total of 100 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Text-Guided Multi-frequency Latent Diffusion for Medical Image Segmentation

    by Qiang Gao, Yi Wang, Yong Zhang, Lan Du, Yong Li, Cunjian Chen

  • A lightweight scale-view co-awareness network for 3D brain tumor segmentation

    by Yong Zhang, Cunjian Chen, Qiang Gao, Yi Wang, YongBing Deng, Bin Fang

  • Consistent and Controllable Image Animation With Linear Motion Diffusion Transformers

    by Xin Ma, Yaohui Wang, Genyun Jia, Xinyuan Chen, Tien-Tsin Wong, Cunjian Chen

    Image animation has seen significant progress, driven by the powerful generative capabilities of diffusion models. However, maintaining appearance consistency with static input images and mitigating abrupt motion transitions in generated animations remain persistent challenges. While text-to-video (T2V) generation has demonstrated impressive performance with diffusion transformer models, the image animation field still largely relies on U-Net-based diffusion models, which lag behind the latest T2V approaches. Moreover, the quadratic complexity of vanilla self-attention mechanisms in Transformers imposes heavy computational demands, making image animation particularly resource-intensive. To address these issues, we propose MiraMo, a framework designed to enhance efficiency, appearance consistency, and motion smoothness in image animation. Specifically, MiraMo introduces three key elements: (1) A foundational text-to-video architecture replacing vanilla self-attention with efficient linear attention to reduce computational overhead while preserving generation quality; (2) A novel motion residual learning paradigm that focuses on modeling motion dynamics rather than directly predicting frames, improving temporal consistency; and (3) A DCT-based noise refinement strategy during inference to suppress sudden motion artifacts, complemented by a dynamics control module to balance motion smoothness and expressiveness. Extensive experiments against state-of-the-art methods validate the superiority of MiraMo in generating consistent, smooth, and controllable animations with accelerated inference speed. Additionally, we demonstrate the versatility of MiraMo through applications in motion transfer and video editing tasks.


  • Open Access

    GlassSplat: Geometric Consistency and Pruning for Reflection-Free 3D Scene Reconstruction

    by Jingjiao You, Yuanyang Zhang, Yining Xu, Yunjie Zhang, Li Yao, Cunjian Chen, Tien-Tsin Wong

    Rendering high-fidelity 3D scenes is crucial for immersive applications like virtual reality and digital twins. However, standard 3D Gaussian Splatting (3DGS) relies heavily on multi-view consistency, making it fragile in real-world scenarios plagued by glass reflections. These reflections often manifest as geometric "floaters" or severe texture artifacts, obscuring the true background. Existing solutions, which typically employ single-image priors or NeRF-based in-painting, often lack explicit 3D constraints or rely on synthetic data, failing to generalize to complex environments. To address these challenges, we first present a novel benchmark dataset of 8 real-world scenes, capturing physically paired reflective and reflection-free images. Building on this, we propose GlassSplat, a robust framework designed to eliminate view-dependent artifacts and recover clean transmission geometry. Our method initializes with a reflection prior and introduces an Affine-Based Exposure Correction module to align global photometric inconsistencies. To distinguishing valid geometry from virtual outliers, we incorporate an Epipolar Consistency Loss and an uncertainty-weighted Depth Regularization. Finally, to physically purge residual noise, we devise a Visibility-Aware Pruning strategy that dynamically filters artifacts based on multi-view statistics. Extensive experiments demonstrate that GlassSplat significantly outperforms state-of-the-art approaches, effectively recovering a clean, artifact-free 3D scene representation.

  • Logits-Level Balanced Machine Unlearning for LLM-Based Recommendation System

    by Chenchen Tan, Xinghao Li, Youyang Qu, Cunjian Chen, Shujie Cui, Longxiang Gao

    As recommendation systems have become foundational to digital platforms, the integration of large language models (LLMs) into these systems offers transformative potential. In LLM-based recommendation (LLMRec) systems, LLMs excel at capturing nuanced user pReferences, greatly enhancing personalization and recommendation accuracy. However, LLMRec systems face data governance issues, such as data privacy, outdated data, poisoned data, and copyrighted data. All these data government scenarios require the removal of data and its impact from large models like LLMRec. This poses significant challenges as existing solutions struggle to accurately delete target data from such complex systems. Motivated by this, we propose an LLMRec unlearning system based on adapter-driven logits modification in this work. This mechanism addresses the unique characteristics of LLMRec’s complex parameters and network structure, enabling precise and effective unlearning while maintaining recommendation performance. Our approach offers several key advantages: 1) using adapters reduces training costs during the unlearning process; 2) a knowledge distillation (KD)-powered logits modification mechanism ensures that the model can still engage in effective reasoning after unlearning, leveraging its existing knowledge logic to maintain high-quality recommendations; and 3) an additional adapter-based general knowledge retention module is incorporated to preserve the core language capabilities of the LLMRec system, ensuring that its foundational performance remains intact. Extensive experimental results show the effectiveness and efficiency of the proposed system. The code is released at: https://anonymous.4open.science/r/llmrecunlearning-1CB3


  • Open Access

    AsyncDiff: Asynchronous Timestep Conditioning for Enhanced Text-to-Image Diffusion Inference

    by Longhuan Xu, Feng Yin, Cunjian Chen

    Text-to-image diffusion inference typically follows synchronized schedules, where the numerical integrator advances the latent state to the same timestep at which the denoiser is conditioned. We propose an asynchronous inference mechanism that decouples these two, allowing the denoiser to be conditioned at a different, learned timestep while keeping image update schedule unchanged. A lightweight timestep prediction module (TPM), trained with Group Relative Policy Optimization (GRPO), selects a more feasible conditioning timestep based on the current state, effectively choosing a desired noise level to control image detail and textural richness. At deployment, a scaling hyper-parameter can be used to interpolate between the original and de-synchronized timesteps, enabling conservative or aggressive adjustments. To keep the study computationally affordable, we cap the inference at 15 steps for SD3.5 and 10 steps for Flux. Evaluated on Stable Diffusion 3.5 Medium and Flux.1-dev across MS-COCO 2014 and T2I-CompBench datasets, our method optimizes a composite reward that averages Image Reward, HPSv2, CLIP Score and Pick Score, and shows consistent improvement.

  • Navigating Unlearning in Medical AI: A Framework for Diabetic Retinopathy Classification

    by Xinghao Li, Chi Liu, Youyang Qu, Shujie Cui, Cunjian Chen, Xingliang Yuan, Zongyuan Ge, Longxiang Gao

    Medical Artificial Intelligence (AI) systems hold great potential for clinical and diagnostic applications. However, their deployment raises significant privacy and ethical concerns, primarily due to the reliance on large-scale datasets containing sensitive health information. The processes of collecting, storing, and using patient data bring forth widespread concerns about privacy protection and data ownership. Additionally, the dynamic nature of medical data presents risks related to outdated or inaccurate information, which can directly impact patient care. Introducing the concept of “unlearning” in medical AI is therefore crucial, not only to address patient privacy needs but also to ensure that models remain adaptable to evolving medical knowledge and data quality. Nevertheless, implementing unlearning in medical AI is considerably more complex compared to other domains, as errors in this context could lead to severe consequences for patient health. This paper aims to establish an evaluation framework specifically for unlearning in classification tasks on the Mobile Brazilian Retinal Dataset (mBRSET) in medical AI. The framework addresses key issues such as privacy, clinical safety, diagnostic accuracy, and fairness. By providing a detailed foundation for assessing unlearning practices, the framework is tailored to the unique challenges presented by the mBRSET dataset. Ultimately, this work seeks to identify the trade-offs and requirements necessary to implement responsible unlearning processes in this highly sensitive domain, thereby contributing to the safe and ethical advancement of medical AI.

  • A Gray-Box Attack Against Latent Diffusion Model-Based Image Editing by Posterior Collapse

    by Zhongliang Guo, Chun Tong Lei, Lei Fang, Shuai Zhao, Yifei Qian, Jingyu Lin, Zeyu Wang, Cunjian Chen, Ognjen Arandjelović, Chun Pong Lau

    Recent advancements in Latent Diffusion Models (LDMs) have revolutionized image synthesis and manipulation, raising significant concerns about data misappropriation and intellectual property infringement. While adversarial attacks have been extensively explored as a protective measure against such misuse of generative AI, current approaches are severely limited by their heavy reliance on model-specific knowledge and substantial computational costs. Drawing inspiration from the posterior collapse phenomenon observed in VAE training, we propose the Posterior Collapse Attack (PCA), a novel framework for protecting images from unauthorized manipulation. Through comprehensive theoretical analysis and empirical validation, we identify two distinct collapse phenomena during VAE inference: diffusion collapse and concentration collapse. Based on this discovery, we design a unified loss function that can flexibly achieve both types of collapse through parameter adjustment, each corresponding to different protection objectives in preventing image manipulation. Our method significantly reduces dependence on model-specific knowledge by requiring access to only the VAE encoder, which constitutes less than 4% of LDM parameters. Notably, PCA achieves prompt-invariant protection by operating on the VAE encoder before text conditioning occurs, eliminating the need for empty prompt optimization required by existing methods. This minimal requirement enables PCA to maintain adequate transferability across various VAE-based LDM architectures while effectively preventing unauthorized image editing. Extensive experiments show PCA outperforms existing techniques in protection effectiveness, computational efficiency (runtime and VRAM), and generalization across VAE-based LDM variants. Our code is available at https://github.com/ZhongliangGuo/PosteriorCollapseAttack.


  • Open Access

    MTS-Net: Dual-enhanced positional multi-head self-attention for 3D CT diagnosis of May-Thurner Syndrome

    by Yixin Huang, Yiqi Jin, Ke Tao, Kaijian Xia, Jianfeng Gu, Lei Yu, Haojie Li, Lan Du, Cunjian Chen