Automated Garment Measurement for Fashion Retail
Precise and flexible garment measurements from images - no tape measures, no delays, just fashion - forward automation.
Get startedWhy GarmentIQ Is Smarter Than Your Tape Measure?
We taught machines how to measure clothes like your grandma (but faster).
We've set up shop in the digital world, powered by computer vision magic and a fancy mix of models like tinyViT, BiRefNet, and HRNet. Our system removes backgrounds, classifies garments, and figures out where the sleeve ends and the waist begins—all with scary-good precision.
We simplify expert know-how into something machines can understand.
No need for complex coding or tedious formatting. Just select the key points you want to measure - shoulder to cuff, hem width, neckline - and GarmentIQ automatically transforms those inputs into a machine-friendly measurement schema. It's fast, intuitive, and tailored to the way professionals actually work.
It’s flexible, like your favorite pair of stretchy jeans.
GarmentIQ is built to bend to your needs. Want to swap in your own models or tweak the measurement logic for a specific brand or category? No problem. The system is modular and customizable, so you stay in control while still benefiting from a strong out-of-the-box solution.
What Are the Key Features of GarmentIQ?
Garment image classification
Our system accurately classifies garments into categories like tops, trousers, and skirts, ensuring seamless organization.
Garment image segmentation
We use advanced segmentation models to isolate garment features from the background for better measurement accuracy.
Garment measurement instruction generation
Our system generates detailed measurement instructions automatically, tailored to the specific garment type and its characteristics.
Garment landmark extraction
Key landmarks are extracted from garment images, enabling precise measurement locations for consistent results.
Garment landmark adjustment
Landmarks can be refined and adjusted manually to ensure they align perfectly, improving the accuracy of garment measurements.
We're students at ETH Zürich, diving deep into data science.
Contributors of this project (alphabetical order by last name):
- Fangzhou Ma ( )
- Li Yuan ( )
- Xinrui Zhai ( )
Some technical detail
Now, let's dive into some of the key numbers and metrics that drive our system.
Our model consistently achieves over 95% accuracy in garment type classification, ensuring reliable results.
We've trained our system with over 23,000 diverse fashion images for robust model performance.
Predictions are made in under 0.5 second (device dependent), delivering real-time garment measurements in production.
All predictions can run fully on-device, ensuring complete data privacy with no cloud dependency.
Take a look at our documentation!
Explore our comprehensive documentation to understand how GarmentIQ works, integrate it into your system, and customize it for your specific needs. We’ve made it easy to get started!
We extend our appriciation to
