darknet-sys 0.3.0

-sys crate for Rust darknet wrapper
Documentation

Main crate: darknet

darknet-sys: FFI bindings to AlexeyAB's Darknet

Usage

Get the crate by adding the dependency to your Cargo.toml.

darknet-sys = "0.3"

If you clone the repository manually, run git submodule init && git submodule update --recursive to get all submodules.

Cargo Features

  • enable-cuda: Enable CUDA (expects CUDA 10.x and cuDNN 7.x).
  • enable-cudnn: Enable cuDNN
  • enable-opencv: Enable OpenCV.
  • runtime: Link to darknet dynamic library. For example, libdark.so on Linux.
  • dylib: Build dynamic library instead of static
  • buildtime-bindgen: Generate bindings from darknet headers.

Build

By default, darknet-sys compiles and link to darknet statically. You can control the feature flags to change the behavior.

Method 1: Download and build from source (default)

By default, it builds and links to darknet statically.

cargo build

You can optionally enable CUDA and OpenCV features. Please read Build with CUDA section to work to CUDA properly.

cargo build --features enable-cuda,enable-opencv

Method 2: Build with custom source

If you prefer to build with your darknet source, fill the source directory to the DARKNET_SRC environment variable. It expects a CMakeLists.txt in that directory.

export DARKNET_SRC=/path/to/your/darknet/repo
cargo build

Method 3: Link to darknet dynamic library

With runtime feature, darknet-sys will not compile the darknet source code and instead links to darknet dynamically. If you are using Linux, make sure libdark.so is installed on your system.

cargo build --feature runtime

Re-generate bindings

With buildtime-bindgen feature, darknet-sys re-generates bindings from headers. It guesses the header file paths according to feature flags. The option is necessary only when darkent is updated or modified.

If you prefer to your (possibly modified) header files, fill the header directory to DARKNET_INCLUDE_PATH environment variable.

Build with CUDA

Please check both CUDA 10.x and cuDNN 7.x are installed and versions are correct.

They are not installed to standard paths on most systems. Please add the CUDA library directory to LIBRARY_PATH environment variable before building. For example on Ubuntu and CUDA 10.1,

env LIBRARY_PATH=/usr/local/cuda-10.1/lib64
cargo build --features enable-cuda

License

MIT license.

Credits

Huge thanks to jerry73204