Kniesel Hannah, Poonam Poonam, Payer Tristan, Bergner Tim, Hermosilla Pedro, Ropinski Timo
Visual Computing Group, Ulm University, Ulm, Germany.
Central Facility for Electron Microscopy, Ulm University, Ulm, Germany.
J Microsc. 2025 Sep;299(3):287-300. doi: 10.1111/jmi.70005. Epub 2025 Jun 28.
Deep learning (DL) has transformed image analysis, enabling breakthroughs in segmentation, object detection, and classification. However, a gap persists between cutting-edge DL research and its practical adoption in electron microscopy (EM) labs. This is largely due to the inaccessibility of DL methods for EM specialists and the expertise required to interpret model outputs. To bridge this gap, we introduce DeepEM Playground, an interactive, user-friendly platform designed to empower EM researchers - regardless of coding experience - to train, tune, and apply DL models. By providing a guided, hands-on approach, DeepEM Playground enables users to explore the workings of DL in EM, facilitating both first-time engagement and more advanced model customisation. The DeepEM Playground lowers the barrier to entry and fosters a deeper understanding of deep learning, thereby enabling the EM community to integrate AI-driven analysis into their workflows more confidently and effectively.
深度学习(DL)已经改变了图像分析,在分割、目标检测和分类方面取得了突破。然而,前沿的深度学习研究与其在电子显微镜(EM)实验室中的实际应用之间仍然存在差距。这主要是因为深度学习方法对于电子显微镜专家来说难以获取,并且解释模型输出需要专业知识。为了弥合这一差距,我们推出了DeepEM Playground,这是一个交互式、用户友好的平台,旨在让电子显微镜研究人员(无论有无编码经验)都能够训练、调整和应用深度学习模型。通过提供一种有指导的实践方法,DeepEM Playground让用户能够探索深度学习在电子显微镜中的工作原理,促进初次接触以及更高级的模型定制。DeepEM Playground降低了入门门槛,促进了对深度学习的更深入理解,从而使电子显微镜领域的研究人员能够更自信、更有效地将人工智能驱动的分析整合到他们的工作流程中。