README.md

• Getting started

  • Checking out the source code
  • Trying samples (C++, C#, Java and Python)
  • GPGPU acceleration
  • Roadmap
  • Getting help

• Online web demo at https://www.doubango.org/webapps/kyc-documents-verif

• Full documentation for the SDK at https://www.doubango.org/SDKs/kyc-documents-verif/docs

• Supported languages (API): C++, C#, Java and Python

• Open source Computer Vision Library: https://github.com/DoubangoTelecom/compv

---

Documents recognition and verification is the core feature for all KYC (Know Your Customer) solutions.

With support for more than 140+ languages, 250+ countries/territories, 5,000+ (and counting) document formats (Passport, Driver license, ID card, Resident card, Visa...)... we built a solution that can boost your KYC platform and put your company ahead of the competition.

We can automatically determine the issuing country and extract every field from the document (Name, Date of Birth, Date of Expiry, Address, Portrait, Signature...) with zero configuration. We also return the exact location (bounding boxes) of each field for visual inspection and verification.

We use state-of-the-art Deep Learning methods developed using Keras with Tensorflow benkend. Our framework is GPGPU accelerated using CUDA and optimized for CPUs (SIMD) using Intel OpenVINO. On both CPUs and GPUs the result is returned in few milliseconds.

Getting started

This version supports both Windows and Linux x86_64.

Checking out the source code

The deep learning models, the binaries and the datasets are on a private repository for obvious reasons. To get access to the private repository you'll need to:

• 1/ Opt-out from all Doubango's private repositories. You cannot be beta tester on more than 1 repo at the same time.
• 2/ Send us a mail with your company name and Github user name (to be added to the private repo). The mail must come from @YourCompanyName, mails from other domains (e.g. @Gmail) will be ignored. The terms of use do not allow you to decompile or reverse engineer the models.

To checkout the source code: git clone --recurse-submodules -j8 https://github.com/DoubangoTelecom/KYC-Documents-Verif-SDK If you already have the code and want to update to the latest version: git pull --recurse-submodules

GPGPU acceleration

We support CUDA, OpenCL and OpenVINO GPGPU acceleration. More info here

Roadmap

This is not a wish list, we already have all listed features already working on demos.

• Image forgery detection
• Barecode reader
• Document liveness detection (will be based on https://github.com/DoubangoTelecom/FaceLivenessDetection-SDK)
• Credit card OCR (improved version of https://github.com/DoubangoTelecom/ultimateCreditCard-SDK)
• Bank check information extraction/OCR from Magnetic Ink Character Recognition [MICR] (E-13B & CMC-7) (improved version of https://github.com/DoubangoTelecom/ultimateMICR-SDK)
• Age estimation from the portrait

Trying samples (C++, C#, Java and Python)

Go to the samples folder and choose your prefered language. The binaries folder contains pre-built bins of the samples which means you don't need to build the samples to try them. The Verify sample is a command line application accepting an image (jpeg, png, bmp...) as input and returning the result as JSON string. The JSON string is the same as what is returned by the online demo hosted at https://www.doubango.org/webapps/kyc-documents-verif.

• Windows: pre-built binary at binaries/windows/x86_64/verify.exe. You can also use the bat file at binaries/windows/x86_64/verify.bat to make your life easier.
• Linux: pre-built binary at binaries/linux/x86_64/verify. You can also use the sh file at binaries/linux/x86_64/verify.sh to make your life easier.

Use the benchmark application to check how fast the SDK can run on your machine.

Technical questions

Please check our discussion group or twitter account