Skip to main content


Cvitek provides TDL integration algorithms to reduce the time required for application development.

This architecture realizes the algorithm required by TDL, including its pre and post processing, and provides a unified and convenient programming interface.

At present, TDL SDK includes motion detection, face detection, face recognition, face tracking, pedestrian detection, semantic segmentation, license plate recognition, license plate detection, live recognition, IR live recognition, infant detection, cry detection, attitude detection, gesture detection, Gesture Recognition and other algorithms.


Chinese Version(中文版)格式English VersionFormat
深度学习SDK软件开发指南htmlpdfTDL SDK Software Development Guidehtmlpdf
YOLO系列开发指南htmlpdfYOLO Development Guidehtmlpdf


The TDL-SDK program needs to be cross-compiled in the Linux environment of the PC host, such as Ubuntu 22.04 system.

  1. Download toolchain


    Extracting Toolchain:

    tar xvf host-tools.tar.gz

    Enter the toolchain directory and export the toolchain path to the environment variable:

    cd host-tools
    export PATH=$PATH:$(pwd)/gcc/riscv64-linux-musl-x86_64/bin

    Verify that the toolchain is available:

    riscv64-unknown-linux-musl-gcc -v

    The version information of the cross-compilation tool chain can be displayed normally, that is, the toolchain is available:

    $ riscv64-unknown-linux-musl-gcc -v
    Using built-in specs.
    Thread model: posix
    Supported LTO compression algorithms: zlib
    gcc version 10.2.0 (Xuantie-900 linux-5.10.4 musl gcc Toolchain V2.6.1 B-20220906)
  2. Compile cvitek-tdl-sdk

    Download the TDL-SDK source code, the Duo and Duo256M/DuoS repositories are different:

    • Duo:

      git clone
      cd cvitek-tdl-sdk-cv180x
    • Duo256M 和 DuoS:

      git clone
      cd cvitek-tdl-sdk-sg200x

      If the repositories code is updated, you can execute git pull in the code directory to pull the latest code, such as Duo256M and DuoS:

      cd cvitek-tdl-sdk-sg200x
      git pull

      After pulling the latest code, continue compiling.

    Compile samples:

    cd sample

    The generated program is in the corresponding subdirectory in the sample directory. For example, the face detection example sample_vi_fd is located in



    ./ clean

Example description

For detailed descriptions and running methods of each example, please refer to the following chapters.

  • carbonfix