Muslin

Muslin's fit comparison preview

Inspiration

Buying used clothes online should be an easy way to save money and keep clothes in circulation. The uncertainty is fit. A size label tells you little about the cut, and two sellers can measure the same garment differently. I wanted a comparison grounded in a piece of clothing you already know fits.

What it does

Muslin lets a buyer compare a listing's pit-to-pit, length, shoulder, and sleeve measurements with a reference garment. Edit the listing measurements and the differences update immediately. Switch the reference garment to see how the same listing compares with another fit. A measurement preview shows the intended photo workflow and labels its sample values clearly.

How I built it

The interface is React, TypeScript, and Vite. Its measurement engine is a separate TypeScript pipeline: find a Letter or A4 sheet for scale, correct perspective, segment the garment, locate landmarks, and estimate dimensions with uncertainty. The live interface currently uses a seeded preview; its photo analysis engine is not connected to that screen. The fit comparison is interactive and runs in the browser without an account or server.

Challenges and what I learned

The hard part is treating a photo measurement as an estimate. Camera angle, the paper reference, fabric edges, and seller technique can all move the result. I split the geometry and image processing into small functions and kept uncertainty visible instead of presenting a false exact answer. I also learned to give judges a reliable sample path while being explicit about what remains to be integrated.

Business model and scale

The proposed buyer tool is free. Sellers who list many garments could pay for batch measuring, reusable listing templates, and exports. An illustrative $12/month seller plan with an assumed $1/month in incremental hosting and support costs leaves $11/month before acquisition and fixed costs; those assumptions still need customer validation. Since comparison runs locally, serving additional buyers does not require a per-comparison cloud inference call. A production seller service would need storage, authentication, support, and billing.

What's next

Connect the existing photo pipeline to image upload in the interface; test it against varied real garments, backgrounds, and camera angles; let buyers save their own reference garments; then pilot seller tools with a small group of secondhand shops.

Important limits

Muslin does not guarantee that a garment will fit. Stretch, drape, body shape, alteration, and inaccurate seller measurements remain outside the current comparison. The displayed photo measurements and closet items are samples in this submission build.