Build an AI-Powered spell check for design system with Figma & TensorFlow
summary
An open source Figma-ML plugin uses TensorFlowJS to detect UI elements in-browser and then compare detections to Figma library metadata, flagging misaligned or nonconforming components. The author trains models with transfer learning on over 10,000 mobile screenshots and 50,000 annotated bounding boxes, chooses a tuned TF model for real-time accuracy, and adds a Zeplin integration that posts checkpoint comments. Notes cover data sourcing, model selection and fine‑tuning, UI matching logic, and deployment tradeoffs.
AI-written summary · about
chapters (7)
- 0:00Start
- 6:47Dataset
- 8:08ML Model
- 8:54Challenges
- 9:17How to compare?
- 12:14Reactions
- 13:33Improve accuracy
related talks
17mNew options for accessible design
25mUsing Figma to go beyond Figma
29mAre Design Systems Ready for the AI Revolution?
14mEmpowering designers: AI, design systems, and quality
34mFrom Figma to Quality Code: Code Connect + AI
12mThe future of design systems and AI
45mFigma Mockup to Server-driven UI
18mAI for UI