发表机构
Lomonosov Moscow State University; DAIMLD; Innopolis University; FusionBrain Lab（莫斯科国立罗蒙诺索夫大学; DAIMLD; 因诺波利斯大学; FusionBrain实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对现有CAD逆向工程AI方法一次性输出完整程序的不足，提出逐步操作的CADENA模型，推出CADENA-Bench基准，在多数据集上性能优于现有方法，相关资源可公开获取。
AI 中文摘要
计算机辅助设计（CAD）是现代工程的基础，但将现有形状转换为可编辑模型仍需大量专业人员的工作。大多数AI系统一次性输出完整的CAD程序，从不检查中间几何结构。相比之下，人类工程师会逐个特征构建零件，在每次操作后检查剩余待建模内容。我们引入CADENA（西班牙语意为“链”），该模型将3D网格重建为参数化CAD程序，每次逐步增加操作序列，并在每一步将目标与当前预测几何结构进行比较。我们还解决了机械零件逆向工程方法评估基准缺失的问题，推出CADENA-Bench，该基准可衡量不同类别机械零件的性能。CADENA在CADENA-Bench、DeepCAD、Fusion 360和MCB数据集上的表现均优于现有方法。代码、模型权重和CADENA-Bench均可通过对应链接获取。
英文摘要
Computer-Aided Design (CAD) underpins modern engineering, yet converting existing shapes into editable models still demands substantial expert effort. Most AI systems emit the entire CAD program in a single pass, never inspecting the intermediate geometry. In contrast, human engineers build a part feature by feature, checking after each operation what remains to be modeled. We introduce CADENA (Spanish for "chain"), a model that reconstructs a 3D mesh as a parametric CAD program, growing its sequence of operations one at a time and comparing the target with the currently predicted geometry at every step. We also address the lack of benchmarks for evaluating reverse-engineering methods on mechanical parts, introducing CADENA-Bench, a benchmark that measures performance across categories of mechanical parts. CADENA outperforms prior methods on CADENA-Bench and on the DeepCAD, Fusion 360, and MCB datasets. Code is available at https://github.com/zhemdi/cadena, model weights at https://huggingface.co/kulibinai/cadena, and CADENA-Bench at https://huggingface.co/datasets/kulibinai/cadena-bench.
CommentsCode: https://github.com/zhemdi/cadena Model: https://huggingface.co/kulibinai/cadena Benchmark: https://huggingface.co/datasets/kulibinai/cadena-bench
