Jul 7, 2025
Using AI to Automate Custom Tool Storage
A fast, CAD-free method to generate true-to-scale DXF and 3D models for tool storage using just a smartphone and AI. Here’s how we used computer vision and Meta AI’s SAM to solve a real CNC workflow bottleneck.
AI
Automation
Tool Storage
The Bottleneck
Tool handling is one of the most deceptively complex aspects of CNC automation. Over the past six years, we’ve designed and deployed systems covering the entire CNC tool lifecycle — central storage, spindle loading, presetting, washing, balancing. Across all of it, one seemingly simple question keeps surfacing:
How do we store parts in a way that’s both robot-accessible and human-friendly?
For standardized toolholders and collets, the answer is straightforward. But once you move into cutting tools and inserts — with over 1,000 SKUs in a typical shop and constant additions — the complexity scales fast. These parts vary in shape, size, and geometry, and in many cases, there’s no existing 3D model to start with.
Designing custom-fit trays or holders becomes a bottleneck. It’s slow, manual, and prone to error.
So we built a solution — a lightweight, AI-driven workflow to convert any physical part into a 3D-printable organizer using just a smartphone and a plain sheet of paper.
The Workflow
At the core of this workflow: Computer Vision + Meta AI’s Segment Anything Model (SAM 2)
We developed a compact toolchain that enables rapid generation of 1:1 DXF profiles and 3D-printable holders using only a smartphone and a blank A4 sheet.
Workflow Overview:
Image Acquisition
A top-down photo of the part is taken on a white A4 sheet (used as a known reference dimension).Segmentation & Contour Extraction
We use Meta AI’s Segment Anything Model (SAM 2) to perform zero-shot segmentation and isolate the object’s outer contour — works great even for fine geometries.Edge Detection & Perspective Correction
OpenCV handles preprocessing: adaptive thresholding, contour simplification, and correction of skewed angles using homography estimation.DXF + STEP Generation
The contour is converted into a true-to-scale DXF file and passed into a lightweight Python CAD kernel to generate a 3D model (STEP/STL) of a custom tray pocket.Output
The result: a printable insert or tray that securely stores the part — sized accurately to within ±0.5 mm, with no manual modeling involved.
How it helped us
Application: Organizing CNC Tool Drawers at Scale
This project found a useful application in helping us solve a real pain point: organizing our messy CNC tool drawers. We wanted to create custom-fit inserts without manually tracing or modeling every part. With this workflow, the process became remarkably simple. And this is just one example of how the approach can streamline physical-to-digital workflows — with plenty more use cases waiting to be explored.
Final Thoughts
This was a fun proof of concept — combining AI and simple CV tools into a real-world, time-saving workflow. For teams working with small-batch or on-demand production, this method offers a fast way to go from physical parts to digital assets. We’ve already found real utility in the workshop, and we’re open to exploring more use cases.
Have an idea where this could be useful?
If you’ve ever needed a fast way to go from physical object to digital outline — this might save you hours. We’d love to hear how you’d apply it — or even collaborate to expand it. Reach out and let us know!
