Identify clothes, then actually shop them.

What a detector can and cannot read from a garment

Identifying clothes is a layered problem. The shallow layer is category: a detector can tell a trench coat from a bomber, a loafer from a derby, a tote from a crossbody, and it does so from silhouette, proportion and how the fabric falls. That layer is quick and rarely wrong on a clear photo.

The deep layer is identity. Whether this trench is a particular product from a particular brand depends on small evidence: the shape of the storm flap, the lining color showing at the cuff, the buckle hardware, a monogram on a button, or a logo that is only a few pixels wide in a screenshot. Much of that evidence is missing or blurred in a real-world photo, so identification stops at a strong guess. Remode marks the difference plainly. Only when the evidence supports the same product does the result say exact; a good-looking guess is labeled best visual match, and a category-level result is labeled similar.

Identifying a piece from a paused video

1. Play the video and pause on the frame where the garment is largest and least blurred, usually a moment when the person stands still or turns slowly.
2. Take a screenshot on your iPhone, then check it in Photos: if the piece looks soft or is half out of frame, go back and pause on a neighboring frame.
3. Open Remode and upload the screenshot. The pieces in the frame are marked, and you tap the one you want identified.
4. Look at what the detector recognized. If it has merged a shirt and a cardigan into one region, crop the screenshot to the piece and try again.
5. Read the label on the result and open the store page to compare the details your frame shows clearly.
6. Save the piece if you are not ready to buy; Remode watches it for price drops, restocks and second-hand listings at no cost.

When identified is not the same as confirmed

For a recognition task the label is the whole answer. Exact tells you the identification is complete: brand, product and, as far as the listing shows, the variant. Best visual match tells you the category and style were identified but the product was not confirmed; treat it as a lead, and check the store page for the detail your photo could not show, such as the back of the garment or the inside label. Similar tells you identification reached the category and stopped there.

Before opening a store, look at the region the result is routed to, because the same piece can carry a different name or price by market. Also note whether the listing is new or second-hand, since a pre-owned listing may be an older version of the same design.

Giving the detector more to work with

• Choose the frame with the garment lit from the front. Backlight turns a textured knit into a flat silhouette and hides the weave that helps identify it.
• Include the label-bearing zone when you can: the chest of a sweatshirt, the back pocket of jeans, the tongue of a sneaker, the clasp of a bag.
• Avoid mirror selfies with reversed text if a plain photo exists; a mirrored logo is harder to read.
• Search jewelry and watches from the closest frame available, since their identifying details are small.
• Skip beauty filters and skin smoothing; they blur fabric texture along with skin.

Pieces that cannot be identified from a photo

Some garments carry no identity to recover. Unbranded basics from a market, a hand-knitted sweater, a vintage piece with the label cut out or a garment altered beyond its original cut can be categorized but not identified, and the result will say similar. Tiny brands that sell through direct messages or a stall, without product pages, leave nothing for the search to confirm. Heavily edited or generated images, extreme compression and frames where the piece is mostly hidden defeat detection outright. For all of these, the realistic path is to shop the similar results or to save the piece and let Remode watch for a listing to appear.