Cover Cropper · Covers + Belts

Drop photos here

Or click to choose files · JPG, PNG · any size

Choose files

Engine

Idle — model loads on first process

ModelRMBG-1.4

Backend—

Settings

Product Output naming JPEG quality

How it works

Each photo is run through a segmentation model (RMBG-1.4) entirely in your browser. The resulting mask gives a precise outline of the product — no shadow inflation, no missed white bands.

Headcovers: classified as upright (front/back) or side. Upright shots are centered and scaled. Side shots split into TOP + BOTTOM. Set each photo's size to driver, fairway, hybrid or putter.

Blade and mallet putter covers are sized separately, each calibrated against its approved reference crop, and neither is ever split into TOP/BOTTOM. Height drives the scale because it barely changes when a cover is turned from face-on to three-quarter. A cover lying on its back is detected automatically and sized to width instead, lifted slightly so it doesn't read as bottom-heavy.

Use Set all above the queue to size a whole shoot in one go; already-finished photos re-render at the new size.

Hats and towels get one crop each with no shot types to pick. Both are matched on width — the live site frames a visor and a snapback to the same width even though they're nowhere near the same height — and centred rather than sat on a baseline. Like belts, they are pure crops: the subject is never scaled up against filled background.

Belts: classified as wound up (coiled roll), flat (full belt, far), or close-up (buckle / interior — subject runs off the frame edges). Every belt output is a pure crop of the source photo — the subject is never scaled up against filled background, so centering yields slightly if the source framing requires it.

Specs

Output1050 × 1400

FormatJPEG

Mask1024 px input

ResampleBicubic

If a crop looks off

Click Re-classify on any job to manually override the detected shot type, or flip the orientation if TOP/BOTTOM came out swapped. Files are processed entirely on your computer — nothing is uploaded.

The model (~88 MB) downloads once on first use and is cached by the browser. Subsequent runs are instant.

Drop photos here

Photos to process

Re-classify shot