Number plates vs appearance: which sorts race photos better?
Sorting, solved. 30+ photographers · 20,000+ photos a minute · 100% group purity · mAP 0.996 · 5 live stores →
Reading the number works until the number is not there. Packs, corners, wet sessions and practice days without plates are the frames that decide how long your event takes, and they are the ones a digit cannot help with.
Both approaches end at the same place, a folder per rider, but they start from different things. One starts from the digits on a board. The other starts from the rider and the machine. That difference decides what happens to the frames where the digits are not there, which on a real trackday is most of the interesting ones. See the whole workflow, from the shoot to the sale.
Number recognition reads the digits on a bike or a bib. RaceLabs groups by rider and vehicle signatures: the leathers, the helmet, the machine and how they go together. The number is still useful, but as a label rather than a key. It names the group once the photographs are already in it.
What number recognition does
Optical character recognition reads visible digits, and when the plate is square to the lens and in focus it reads them well. That is a real strength on a start-line shot or a static portrait, where the number is there to be read.
The catch is what the number refers to. Bikes get shared between sessions, numbers get reused across classes, and a 6 read as an 8 sends a set to the wrong rider without anything looking wrong.
When numbers are hard to read
Now count how much of your card has a legible plate in it. Packs, corners, pans and wet sessions are where the good frames live, and they are exactly where the number is hidden, angled away or smeared.
- Packs. The plate is hidden behind another bike. Nothing to read.
- Cornering. Mid-corner the number is angled away from the lens. No readable digits.
- Motion blur. A pan on the straight smears the number. The read may be uncertain or wrong.
- Weather. Rain and mud cover the plate. The number may be unreadable.
- No numbers at all. Free practice, testing and many trackdays run with no numbers fitted. There is no number to read.
- Full-face helmets. A concealed face cannot provide visible facial detail. Check what other information the workflow can use.
With no legible number there is nothing for a reader to work with, and those photographs land in a pile you sort by hand. That pile, not the easy frames, is what sets how long your event takes.
How appearance matching is different
A signature is built from what stays the same all day: the leathers, the helmet design, the machine and the way the rider sits on it. It survives the rear view, the pack and the pan, so those frames land in the right rider without a number anywhere in shot. Similar equipment in the same class is the hard case, and it is the one that goes to the Unclear bin rather than to the wrong rider.
RaceLabs’ October 2026 mixed benchmark covers 175 held-out identities and 6,504 photographs. It reports retrieval mAP 0.996 and 100% group purity: not one of 6,504 photos in the wrong rider’s group. See the published validation and the motorcycle rider-group examples.
A readable number can help name a rider group. Inspect split groups and decide which need merging; a number does not replace the review of the photographs.
Side by side
| Real race-weekend frame | Number recognition alone | Appearance matching (RaceLabs) |
|---|---|---|
| Plate hidden in a pack | No digits to read | Grouped by rider signature |
| Mid-corner, number angled away | No digits to read | Connected across views of the same rider |
| Motion blur at speed | Digits unreadable | Placed when the signature holds, sent to Unclear when it does not |
| Practice or testing, no numbers | Nothing to read | Grouped without a number at all |
| Rain and mud over the plate | No digits to read | Grouped by rider signature |
The cost of a wrong match
A rider expects the gallery to contain their photographs. In the published benchmark, B³ precision measures group purity: how much of a rider’s gallery really is that rider. Before you publish, glance over the groups and merge any rider split in two.
The benchmark reports 100% group purity: not one of 6,504 photos landed in the wrong rider's group. That is the number that matters, because a rider finding someone else in their gallery is a refund, while merging two groups is one click.
So which should you use?
If every rider on your card shows a clean plate to the lens, reading the number is enough, and a start-line shoot is exactly that. A trackday is not: the frames people buy are the ones taken mid-corner, in a pack or in the rain, and a number cannot find those. That is where a signature does the work.
RaceLabs keeps your originals on your disk during sorting and sends lightweight previews for cloud analysis. Include those stages and your corrections when timing a complete event from preparation to publication.
Follow the trackday sorting workflow, finish your edits and prepare the customer gallery. A connected RaceLabs photographer website lets you upload finished full-resolution photos and videos, then publish the event under your own brand.
The trackday-photo selling guide covers offers, publication and downloads. Explore real photographer stores for the finished customer experience.
Your first event is free.
Bring a complete shoot, keep the originals on your disk and see the rider groups before you pay anything.
Frequently asked questions
Can I sort race photos without reading the number plate?
Yes. Rider and vehicle signatures help RaceLabs group photographs across an event without a readable number. A visible number can help name a group. Review uncertain matches and split sets before delivery.
How should I compare sorting tools?
Compare the tools on the same complete event. Number recognition needs readable digits for that part of the task; RaceLabs uses rider and vehicle signatures to help group photographs without a readable number. Check the results and the correction work in both cases.
What happens when appearance matching is unsure?
Uncertain frames go to the Unclear bin instead of into the wrong rider, which is why precision stays high: the October 2026 benchmark reports 100% group purity. Check the bin and merge any split groups before you publish.
Does a readable number still help?
Yes. A readable number can help name a group. Inspect the photographs and decide whether split groups should be merged before delivery.
What does RaceLabs cost?
At the €0.01 top rate, 3,000 photos cost €30. Marginal volume bands go down to €0.004, on a count that resets every year; confirm the applicable schedule in your quote. Your first event is free. Tell us your file mix and volume to agree the setup. See the photo-sorting rates and volume bands.