Custom code recognition development for your task
GS1 DataMatrix, GS1-128 (SSCC), retail barcodes and QR — on your line, with your camera, in your system. We start with a free assessment on your frames: you send images, we tell you how many read today and what can be improved.
What we do
- Reading on your frames. Small modules, conveyor motion blur, film glare, curved packaging, faint print — we tune and extend the reader for your images, not for textbook ones.
- A new code or marking format. Another symbology, your own string format, special rules for validating code content.
- Integration into your system. ERP, MES, WMS or your own service: C++ and C ABI, Python, a command-line tool with a service mode. We help connect it and test on your setup.
- Builds for your processor and environment. x86-64 with AVX2 and AVX-512, ARM64 and ARMv7 with NEON (including Raspberry Pi), Windows and Linux, Docker and Kubernetes, offline machines.
- Per-frame deadline. A guaranteed response time per frame matched to your line speed: the library stops searching at the deadline and returns what it found.
- Camera and lighting. We advise on resolution, field of view and exposure for your module size — before you buy a camera.
How we work
- You send frames and the task. 50 to 200 images from the line or warehouse, as they are, including failed ones; a few words on the task: which codes, what speed, where results go. Large archives — as a cloud drive link.
- Free assessment. We run our readers on your frames and send a report: how many codes read, how many do not and why (module size, blur, glare, damage), what development would give and what a camera or lighting change would.
- Estimate and timeline. Price and timeline are quoted after the assessment: they depend on the task, not on a price list.
- Work and acceptance on your frames. Acceptance is measured on your set: read rate, false reads, time per frame.
- Delivery and support. A build for your platform, documentation, help with the rollout, and updates.
What we have already built for ourselves
Our libraries and services — the same kind of work we will do on your task:
- Conveyor DataMatrix reader — reads symbols at about 2 pixels per module; 99.6% of production-line frames, no false reads, 7 ms median per frame.
- Barcode reader — EAN/UPC, Code 128 and GS1-128 (SSCC) — 72% of symbols on public datasets vs 54–56% for zxing-cpp and ZBar, no false reads on images without a code.
- Scout network — a small neural network inside the library finds codes in the frame before reading, so several codes read faster and more completely.
- 20–40 MP camera frames — reading small codes on large frames from modern cameras.
- NEON and AVX-512 builds — kernels for ARM and modern x86: one library on a Raspberry Pi and on a server.
- DataMatrix generator — with GS1 and Chestny ZNAK profile validation; every symbol is read back before it is returned.
- Chestny ZNAK status check — code status via the True API in the online decoder.
- Messenger bot — decodes DataMatrix from a photo right in a MAX chat.
Numbers and methodology: the comparison with zxing-cpp, libdmtx and ZBar.
Discuss your task
Tell us which codes you read and where, and attach frames or a link to them.
Or by email: libscanner@yandex.com