Making Training-Free Diffusion Segmentors Scale with the Generative Power
Institute of Information Engineering, CAS（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络空间安全学院）；State Key Laboratory of AI Safety, Institute of Computing Technology, CAS（中国科学院计算技术研究所人工智能安全国家重点实验室）；School of Cyber Science and Tech., Shenzhen Campus of Sun Yat-sen University（中山大学深圳校区网络空间安全学院）；Alibaba Group（阿里巴巴集团）；School of Computer Science and Tech., University of Chinese Academy of Sciences（中国科学院大学计算机科学与技术学院）；Key Laboratory of Big Data Mining and Knowledge Management, CAS（中国科学院大数据挖掘与知识管理重点实验室）；Beijing Academy of Artificial Intelligence（北京人工智能研究院）
As powerful generative models, text-to-image diffusion models have recently been explored for discriminative tasks. A line of research focuses on adapting a pre-trained diffusion model to semantic segmentation without any further training, leading to training-free diffusion segmentors. These methods typically rely on cross-attention maps from the model's attention layers, which are assumed to capture semantic relationships between image pixels and text tokens. Ideally, such approaches should benefit from more powerful diffusion models, i.e., stronger generative capability should lead to better segmentation. However, we observe that existing methods often fail to scale accordingly. To understand this issue, we identify two underlying gaps: (i) cross-attention is computed across multiple heads and layers, but there exists a discrepancy between these individual attention maps and a unified global representation. (ii) Even when a global map is available, it does not directly translate to accurate semantic correlation for segmentation, due to score imbalances among different text tokens. To bridge these gaps, we propose two techniques: auto aggregation and per-pixel rescaling, which together enable training-free segmentation to better leverage generative capability. We evaluate our approach on standard semantic segmentation benchmarks and further integrate it into a generative technique, demonstrating both improved performance broad applicability. Codes are at https://github.com/Darkbblue/goca.
Generator Versus Segmentor: Pseudo-healthy Synthesis
Xiamen University（厦门大学）；Deepwise Inc.（深睿公司）
本文研究了伪健康图像的合成问题，将其定义为从病理图像中合成特定对象的无病理图像。最近的方法基于生成对抗网络（GAN）已经开发了这个任务。然而，这些方法将不可避免地陷入保持对象特定的身份和生成健康的外观之间的权衡。为了克服这一挑战，我们提出了一种新的对抗性训练机制，Generator versus Segmentor（GVS），通过分治策略来减轻这种权衡。我们进一步考虑了在整个训练过程中分割器的泛化性能不断恶化，并通过对经过良好变换的像素进行静音来提高像素加权损失。此外，我们提出了一种新的度量标准来衡量合成图像的健康程度。在公共数据集BraTS上的定性和定量实验表明，该方法优于现有方法。此外，我们还证明了我们的方法在数据集LITS的有效性。我们的实施和预先训练的网络可在www.example.com上公开获取。
Red-GAN: Attacking class imbalance via conditioned generation. Yet another perspective on medical image synthesis for skin lesion dermoscopy and brain tumor MRI
C^2GR: Coupled Comprehensive Generative Replay for a Continually Learnable Universal Segmentation Model
Shanghai Jiao Tong University（上海交通大学）；Southeast University（东南大学）；Stevens Institute of Technology（史蒂文斯理工学院）；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；Imperial College London（伦敦帝国学院）
Medical World Model: Generative Simulation of Tumor Evolution for Treatment Planning
The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；Johns Hopkins University（约翰霍普金斯大学）；The First Affiliated Hospital of Nanjing Medical University（南京医科大学第一附属医院）；University of California, San Francisco（加州大学旧金山分校）
本文提出医学世界模型 MeWM，结合视觉-语言策略模型与肿瘤生成动力学模型模拟治疗后肿瘤演化，并通过逆动力学模型优化个体化治疗方案，将最优 TACE 方案选择的 F1 分数提升 13%。
HiDiff: Hybrid Diffusion Framework for Medical Image Segmentation
Institute of Science and Technology for Brain-inspired Intelligence, Fudan University（脑启发智能科学与技术研究院，复旦大学）；Shanghai Key Lab of Intelligent Information Processing, School of Computer Science, Fudan University（上海智能信息处理重点实验室，复旦大学计算机学院）
SVBRDF Recovery From a Single Image With Highlights using a Pretrained Generative Adversarial Network
Nanjing University of Science and Technology（南京理工大学）；The Hong Kong Polytechnic University（香港理工大学）；Nanjing University（南京大学）；Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK（格勒诺布尔阿尔卑斯大学、法国国家信息与自动化研究所、法国国家科学研究中心、格勒诺布尔理工学院、LJK实验室）
METGAN: Generative Tumour Inpainting and Modality Synthesis in Light Sheet Microscopy
Technische Universität München（慕尼黑工业大学）；Helmholtz Zentrum München（亥姆霍兹慕尼黑研究中心）；TranslaTUM Center for Translational Cancer Research（TranslaTUM转化癌症研究中心）；University of Zürich（苏黎世大学）
Distribution Aligned Diffusion and Prototype-guided network for Unsupervised Domain Adaptive Segmentation
Xi’an Jiaotong University（西安交通大学）；The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；University of California, Santa Cruz（加州大学圣克鲁兹分校）
