第一届面向可控生成的解耦表示学习国际研讨会（DRL4Real）：方法与结果
机器学习
2025-09-16 v1 计算机视觉与模式识别
摘要
本文回顾了与 ICCV 2025 联合举办的第一届面向可控生成的解耦表示学习国际研讨会（DRL4Real）。该研讨会旨在弥合解耦表示学习（DRL）的理论前景与其在现实场景中应用之间的差距，超越合成基准测试。DRL4Real 侧重于在可控生成等实际应用中评估 DRL 方法，并探索模型在鲁棒性、可解释性和泛化性方面的进展。该研讨会接收了 9 篇论文，涵盖广泛的主题，包括新归纳偏置（如语言）的整合、扩散模型在 DRL 中的应用、3D 感知解耦，以及 DRL 向自动驾驶和脑电图（EEG）分析等专业领域的扩展。本摘要详细介绍了研讨会的目标、所接收论文的主题，并概述了作者们提出的方法。
引用
@article{arxiv.2509.10463,
title = {The 1st International Workshop on Disentangled Representation Learning for Controllable Generation (DRL4Real): Methods and Results},
author = {Qiuyu Chen and Xin Jin and Yue Song and Xihui Liu and Shuai Yang and Tao Yang and Ziqiang Li and Jianguo Huang and Yuntao Wei and Ba'ao Xie and Nicu Sebe and Wenjun and Zeng and Jooyeol Yun and Davide Abati and Mohamed Omran and Jaegul Choo and Amir Habibian and Auke Wiggers and Masato Kobayashi and Ning Ding and Toru Tamaki and Marzieh Gheisari and Auguste Genovesio and Yuheng Chen and Dingkun Liu and Xinyao Yang and Xinping Xu and Baicheng Chen and Dongrui Wu and Junhao Geng and Lexiang Lv and Jianxin Lin and Hanzhe Liang and Jie Zhou and Xuanxin Chen and Jinbao Wang and Can Gao and Zhangyi Wang and Zongze Li and Bihan Wen and Yixin Gao and Xiaohan Pan and Xin Li and Zhibo Chen and Baorui Peng and Zhongming Chen and Haoran Jin},
journal= {arXiv preprint arXiv:2509.10463},
year = {2025}
}
备注
Workshop summary paper for ICCV 2025, 9 accepted papers, 9 figures, IEEE conference format, covers topics including diffusion models, controllable generation, 3D-aware disentanglement, autonomous driving applications, and EEG analysis