Our digital twin platform eliminates guesswork, allowing customers to shop only what fits.
Follow Us:
Quick Links
Quick Links
Contact Us
Introduction
How the FIT:MATCH.ai Patented algorithms 'FIT'
into our revolutionary market solution.
At FIT:MATCH.ai, our technological journey starts with a problem, specifically a measurement problem.
The two main methods of extracting body measurements are not accurate and do not translate in a way that correctly measures an individual's unique shape.
Throughout our research, we found issues with both survey-based fit analysis and surrounding body scanning fit analysis.
We’ve concluded that for a 3D problem, we need a 3D solution.
The Fit Challenge
Survey-Based Fit Analysis
The average correlation among 96 dimensions of body size is only 0.43, so someone’s height, neck, waist or other measurements, is unlikely to tell you much about their other precise dimensions which are important for most categories of clothing.
Other challenges include a lack of body data points.
Consumers don’t know their measurements.
Consumers do not want to measure their body or don't have access to a traditional tape measure.
Consumers do not know how to measure their body correctly.
Body shape change leads to measurement change.
Body Scanning Based Fit Analysis
The discrepancy between scan measurements and tailor measurements are due to shape.
2D measurements do not reflect body shape details.
2D cannot translate to 3D.
Our Innovative 3D Solution
On the journey to discover the FIT:MATCH.ai solution, we conducted an immense amount of user testing.
We scanned and fitted thousands of female fit testers of all shapes and sizes. Their breast scans were grouped based on their actual best fitting bra sizes.
These scans were processed via slicing and point reorganizing before they were stored.
Increased User Testing
80% of shoppers were wearing the wrong size bra.
Encoded 3D shape
By reorganizing points on scan surface by horizontal slices and angles
On each slice: 1 point at every 5 degrees (-180 to 0 degrees, 37 points per slice)
Record each point’s distance from the origin (0,0).
These sorted distance values constitute the numeric 3D shape info. All scans are processed in this way to have same number of points, sorted in the same order.
Following this data collection and 3D size chart building, we are ready for size predictions for new consumers. Our algorithms compare the shape of the new consumer with each of the shapes we collected before (fit model scans), and calculate a ‘shape discrepancy’ value for each comparison, where larger value indicates larger shape difference. Finally we identify the fit model scan with the lowest shape discrepancy value against the new consumer, who will be assigned to the same size as the identified fit model.
After correct predictions have been made, the new consumer’s scan will be included in the 3D size chart to facilitate constant improvement of our matching algorithms.
Evaluating Shape Discrepancies
In addition to the overall shape discrepancy value, our technology can visualize the local fit gaps by comparing shapes at different body areas.
Purple area: a new consumer’s torso scan
Black area: the fit twin matched to the consumer’s shape
Red areas: shape bigger than fit model in the size chart—> tight areas
Blue areas: shape smaller than fit model in the size chart—> loose areas
This is how we can evaluate body asymmetry
and account for it when we recommend a size.
Data Feedback
In addition, we provide helpful insights to apparel companies. For example, we can make suggestions of patterns modifications or product grade rules change based on body shape info.
For an established sizing system, we can find the most representative shape for each size, so that companies can optimize their fit models’ hire. To the right is how it’s done:
S3 has the lowest aggregate shape discrepancy score, thus it is the most representative shape for 34C and should be chosen as the fit model for the product design.
= 170.0
= 247.4
= 89.6
MIN
Output: S3
Easy Scale Up
The FIT:MATCH.ai technology can be scaled up quickly and efficiently, by taking the data gathered from one style of bra and translating it to a different style of bra. We can see this exemplified with the Demi bra and the T-shirt bra.
Demi Bra Block
Use same scans for different styles.
No need to rescan unless there are major body shape changes.
T-shirt Bra Block
