mlsanityData to 14 Oct 2026

SketchDNN: Joint Continuous-Discrete Diffusion for CAD Sketch Generation

arXiv 2507.11579 ↗First posted 15 Jul 2025cs.CV cs.LGcited by 5

On the map: Diffusion models → Image generation, text-to-image and editing

Abstract (arXiv)

We present SketchDNN, a generative model for synthesizing CAD sketches that jointly models both continuous parameters and discrete class labels through a unified continuous-discrete diffusion process. Our core innovation is Gaussian-Softmax diffusion, where logits perturbed with Gaussian noise are projected onto the probability simplex via a softmax transformation, facilitating blended class labels for discrete variables. This formulation addresses 2 key challenges, namely, the heterogeneity of primitive parameterizations and the permutation invariance of primitives in CAD sketches. Our approach significantly improves generation quality, reducing Fr'echet Inception Distance (FID) from 16.04 to 7.80 and negative log-likelihood (NLL) from 84.8 to 81.33, establishing a new state-of-the-art in CAD sketch generation on the SketchGraphs dataset.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models. Builds on 28 of its 35 references that are in the map. First posted on arXiv 15 Jul 2025.

Builds on

28 of 35 references.

Closest ancestors · 2

Same direction, newest first

  1. 2023DLT: Conditioned layout generation with Joint Discrete-Continuous Diffusion Layout Transformer7 Mar
  2. 2023PLay: Parametrically Conditioned Layout Generation using Latent Diffusion27 Jan

Other references · 26

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  3. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,058
  4. 2022Scalable Diffusion Models with Transformerscited by 6,054
  5. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,761
  6. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  7. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  8. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,148
  9. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 745
  10. 2021Argmax Flows and Multinomial Diffusion: Learning Categorical Distributionscited by 574
  11. 2021DeepCAD: A Deep Generative Network for Computer-Aided Design Modelscited by 271
  12. 2021Autoregressive Diffusion Modelscited by 210
14 more
  1. 2022SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Controlcited by 194
  2. 2022SkexGen: Autoregressive Generation of CAD Construction Sequences with Disentangled Codebookscited by 125
  3. 2024BrepGen: A B-rep Generative Diffusion Model with Structured Latent Geometrycited by 103
  4. 2021Computer-Aided Design as Languagecited by 82
  5. 2021A Review of the Gumbel-max Trick and its Extensions for Discrete Stochasticity in Machine Learningcited by 82
  6. 2022HouseDiffusion: Vector Floorplan Generation via a Diffusion Model with Discrete and Continuous Denoisingcited by 73
  7. 2021Unleashing Transformers: Parallel Token Prediction with Discrete Absorbing Diffusion for Fast High-Resolution Image Generation from Vector-Quantized Codescited by 71
  8. 2022Latent Diffusion Energy-Based Model for Interpretable Text Modelingcited by 67
  9. 2021Vitruvion: A Generative Model of Parametric CAD Sketchescited by 66
  10. 2023TESS: Text-to-Text Self-Conditioned Simplex Diffusioncited by 65
  11. 2021Engineering Sketch Generation for Computer-Aided Designcited by 49
  12. 2023Permutation Equivariance of Transformers and Its Applicationscited by 20
  13. 2019Invertible Gaussian Reparameterization: Revisiting the Gumbel-Softmaxcited by 18
  14. 2022Diffusion bridges vector quantized Variational AutoEncoderscited by 16

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 2.

5 papers.

  1. 2025ReCAD: Reinforcement Learning Enhanced Parametric CAD Model Generation with Vision-Language Modelscited by 11
  2. 2025JointDiff: Bridging Continuous and Discrete in Multi-Agent Trajectory Generationcited by 5
  3. 2025Simplex-to-Euclidean Bijections for Categorical Flow Matchingcited by 3
  4. 2026CADENA: Stepwise CAD Reverse Engineeringcited by 1
  5. 2026Simplex-to-Euclidean Bijection for Conjugate and Calibrated Multiclass Gaussian Process Classificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Computer-Aided Design Generation by Cascaded Discrete Diffusion Model
  2. 2025CoProSketch: Controllable and Progressive Sketch Generation with Diffusion Model
  3. 2025Target-Guided Bayesian Flow Networks for Quantitatively Constrained CAD Generation
  4. 2024Training-Free Sketch-Guided Diffusion with Latent Optimization
  5. 2025SwiftSketch: A Diffusion Model for Image-to-Vector Sketch Generation
  6. 2024KnobGen: Controlling the Sophistication of Artwork in Sketch-Based Diffusion Models
  7. 2023DiffSketcher: Text Guided Vector Sketch Synthesis through Latent Diffusion Models
  8. 2023SketchFFusion: Sketch-guided image editing with diffusion model