AMSnet 2.0
收藏资源简介:
AMSnet 2.0是一个大规模的模拟和混合信号电路设计数据集,包含了2686个电路的原理图图像、Spectre格式的网表、OpenAccess格式的数字原理图以及电路组件和网的位置信息。该数据集是为了解决多模态大语言模型在电路图识别、理解和网表生成方面的困难而创建的。AMSnet 2.0通过一个数据标注平台收集了来自教科书和公开竞赛的原理图图像,并进行了人工标注。数据集的建设过程包括图像收集、原理图元素和网检测以及网表生成。该数据集旨在推动多模态大语言模型在模拟和混合信号电路设计中的应用。
AMSnet 2.0 is a large-scale analog and mixed-signal circuit design dataset. It contains schematic images of 2686 circuits, netlists in Spectre format, digital schematics in OpenAccess format, as well as position information of circuit components and nets. This dataset was developed to address the challenges faced by multimodal large language models (LLMs) in circuit schematic recognition, understanding and netlist generation. AMSnet 2.0 collects schematic images from textbooks and public competitions via a data annotation platform, and conducts manual annotation. The dataset construction process includes image collection, schematic element and net detection, as well as netlist generation. This dataset aims to promote the application of multimodal LLMs in analog and mixed-signal circuit design.

- 1AMSnet 2.0: A Large AMS Database with AI Segmentation for Net Detection宁波数字孪生研究院,东方理工大学,宁波,中国; 加利福尼亚大学洛杉矶分校,美国; 清华大学,北京,中国 · 2025年



