CADENA：逐步CAD逆向工程

摘要

電腦輔助設計（Computer-Aided Design，CAD）支撐著現代工程技術，然而將既有形狀轉換為可編輯模型仍需要大量專家投入。大多數 AI 系統會以單次生成的方式輸出完整 CAD 程式，且從不檢視中間階段的幾何形狀。相較之下，人類工程師會逐個特徵建構零件，並在每次操作後檢查剩餘待建模的部分。我們提出 CADENA（西班牙語「鏈」之意），此模型能將 3D 網格重建為參數式 CAD 程式，以一次生成一個操作的方式逐步增長操作序列，並在每一步將目標與當前預測的幾何形狀進行比較。此外，針對機械零件逆向工程方法缺乏評測基準的問題，我們推出 CADENA-Bench——一個橫跨多個機械零件類別、衡量方法效能的基準測試集。CADENA 在 CADENA-Bench 以及 DeepCAD、Fusion 360、MCB 資料集上均優於先前方法。程式碼請見 https://github.com/zhemdi/cadena，模型權重請見 https://huggingface.co/kulibinai/cadena，CADENA-Bench 請見 https://huggingface.co/datasets/kulibinai/cadena-bench。

English

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.