工业质检领域相关近期顶会论文汇总CVPR2025

异常检测Anomaly Detection

• ⭐️UniNet: A Contrastive Learning-guided Unified Framework with Feature Selection for Anomaly Detection, Code，MVTech当前排名第2
• ⭐️AA-CLIP: Enhancing Zero-shot Anomaly Detection via Anomaly-Aware CLIP, Code
• Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-Class Unsupervised Anomaly Detection, Code
• star:TailedCore : Few-Shot Sampling for Unsupervised Long-Tail Noisy Anomaly Detection, Code
• Odd-One-Out: Anomaly Detection by Comparing with Neighbors, Code

多类别

• One-for-More: Continual Diffusion Model for Anomaly Detection, Code，一个模型多类别异常检测
• AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios, Code
• Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection, Code

数据集

• MANTA: A Large-Scale Multi-View and Visual-Text Anomaly Detection Dataset for Tiny Objects, Dataset

开集有监督异常检测Open-set Supervised Anomaly Detection（OSAD）

• Distribution Prototype Diffusion Learning for Open-set Supervised Anomaly Detection

基于多模态大模型能力

• Towards Training-free Anomaly Detection with Vision and Language Foundation Models, Code

---

骨干网络

Mamba系列（mamba为transformer后的新骨干网络形式）

• MambaVision: A Hybrid Mamba-Transformer Vision Backbone，Code
• MobileMamba: Lightweight Multi-Receptive Visual Mamba Network, Code

其他

• Transformers without Normalization, Code，不用normalization也能取得相同的效果？
• ⭐️ShiftwiseConv: Small Convolutional Kernel with Large Kernel Effect, Code，近期的大卷积核带来了超越transformer的效果，why?
• OverLoCK: An Overview-first-Look-Closely-next ConvNet with Context-Mixing Dynamic Kernels, Code，自上而下的注意力+动态卷积

---

目标检测

开集识别

• LLMDet: Learning Strong Open-Vocabulary Object Detectors under the Supervision of Large Language Models, Code
• SGC-Net: Stratified Granular Comparison Network for Open-Vocabulary HOI Detection, Code

DETR

• DEIM: DETR with Improved Matching for Fast Convergence, Code, 加速DETR类目标检测的训练和推理速度，训练速度减少50%！🙀

实例检测Instance Detection

• Solving Instance Detection from an Open-World Perspective, Code

分割

• Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation，Code
• ROS-SAM: High-Quality Interactive Segmentation for Remote Sensing Moving Object, Code

---

异常生成

• Anomaly Anything: Promptable Unseen Visual Anomaly Generation
• Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation

---

图像编辑

• Edit Away and My Face Will not Stay: Personal Biometric Defense against Malicious Generative Editing, Code
• h-Edit: Effective and Flexible Diffusion-Based Editing via Doob’s h-Transform, Code
• Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing, Code

---

few-shot

视觉语言模型VLM

• Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages, Code
• UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection, Code

迁移学习

• MMRL: Multi-Modal Representation Learning for Vision-Language Models, Code

---

zero-shot

多模态大模型MLLM

• Towards Zero-Shot Anomaly Detection and Reasoning with Multimodal Large Language Models, Code
• Bayesian Prompt Flow Learning for Zero-Shot Anomaly Detection, Code

其他

主动学习

• Rethinking Epistemic and Aleatoric Uncertainty for Active Open-Set Annotation: An Energy-Based Approach, Code，开集主动学习

半监督semi-supervised

• Unlocking the Potential of Unlabeled Data in Semi-Supervised Domain Generalization, Code

增量学习

• Dual Consolidation for Pre - Trained Model - Based Domain - Incremental Learning, Code

参考：

1. https://cvpr.thecvf.com/Conferences/2025/AcceptedPapers
2. https://zhuanlan.zhihu.com/p/30204993641
3. https://github.com/amusi/CVPR2025-Papers-with-Code?tab=readme-ov-file
4. https://github.com/Paper2Chinese/CVPR-2025-reading-papers-with-code?tab=readme-ov-file