Alturos.Yolo

To start using object detection in your C# project, install the core wrapper and a pre-trained dataset via the NuGet Package Manager. For YOLOv2-tiny, you should install both the wrapper and the specific dataset package.

PM> install-package Alturos.Yolo
PM> install-package Alturos.YoloV2TinyVocData

After building darknet inside a Docker container, you can copy the compiled release files to your local host machine using docker cp. Replace <container_id> with your actual running container ID.

# Find your container ID
docker ps

# Copy the build_release folder to a local directory
docker cp <container_id>:/darknet/build_release C:\temp

If you are running Alturos.Yolo within a .NET-based Docker container, you must ensure libgomp1 is installed to support the parallel processing required by the underlying models.

apt-get update
apt-get install libgomp1

To build the darknet dependency required by Alturos.Yolo, use a GCC Docker container to compile it from source. This involves setting up a build environment with necessary libraries (OpenCV, FFmpeg, GTK), cloning the repository, and configuring the Makefile to enable shared libraries (LIBSO=1) and OpenCV support (OPENCV=1).

# 1. Start a GCC container
docker pull gcc
docker run -it gcc /bin/bash

# 2. Install build dependencies
apt-get update && apt-get install build-essential cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev libopencv-dev -y

# 3. Clone and configure darknet
git clone https://github.com/AlexeyAB/darknet.git
cd darknet
sed -i 's/LIBSO=0/LIBSO=1/g' Makefile
sed -i 's/OPENCV=0/OPENCV=1/g' Makefile

# 4. Build
make

The Alturos.Yolo package provides the core logic for CPU-based object detection. For a quick start with pre-trained models, you should also install the Alturos.YoloV2TinyVocData NuGet package.

To get started immediately with object detection, run:

install-package Alturos.Yolo
install-package Alturos.YoloV2TinyVocData

GPU detection is typically 10 times faster than CPU. To enable it, you must satisfy these requirements:

1. Install the latest Nvidia driver.
2. Install Nvidia CUDA Toolkit 10.2.
3. Download Nvidia cuDNN v7.6.5 for CUDA 10.2.
4. Copy cudnn64_7.dll from the cuDNN output directory into your project folder.

By default, Alturos.Yolo uses CPU detection. It will automatically switch to GPU mode if the following Nvidia dependencies are present in your environment:

1. Nvidia CUDA Toolkit 10.2: Must be installed along with a hardware driver that supports CUDA.
2. Nvidia cuDNN v7.6.5 for CUDA 10.2: Specifically, the cudnn64_7.dll file is required.

Setup Step: Copy cudnn64_7.dll (found in the %cudnn%\bin directory) into the x64 directory of your application's output folder.

To set up a development environment for Alturos.Yolo on Linux, use the Microsoft .NET Core SDK Docker image. You will also need to install libgomp1 via apt-get to satisfy dependencies for the object detection logic.

Follow these steps:

1. Pull and run the .NET Core SDK container.
2. Update apt and install vim and libgomp1.
3. Create a new .NET console project.
4. Add the required NuGet packages: Alturos.Yolo and Alturos.YoloV2TinyVocData.

# Pull and enter the SDK container
docker pull mcr.microsoft.com/dotnet/core/sdk
docker run -it mcr.microsoft.com/dotnet/core/sdk /bin/bash

# Inside the container, install dependencies
apt-get update
apt-get install vim -y
apt-get install libgomp1

# Setup the project
mkdir test
cd test
dotnet new console
dotnet add package Alturos.Yolo --version 3.0.3-alpha
dotnet add package Alturos.YoloV2TinyVocData --version 1.0.0

If you encounter errors during runtime, check the following:

• DllNotFoundException: Use the Dependencies tool to verify that all required DLLs (like yolo_cpp_dll_gpu.dll) have their references available.
• NotSupportedException: Ensure you are using the latest Nvidia driver.
• GPU Usage: Use the command %PROGRAMFILES%\NVIDIA Corporation\NVSMI\nvidia-smi.exe to check graphic device usage.

Use ConfigurationDetector to automatically determine the best configuration for your environment, then pass that configuration to a YoloWrapper. The Detect method accepts an image file path and returns a collection of detected items containing the object type, confidence score, and bounding box coordinates.

var configurationDetector = new ConfigurationDetector();
var config = configurationDetector.Detect();
using (var yoloWrapper = new YoloWrapper(config))
{
	var items = yoloWrapper.Detect(@"image.jpg");
	// items[0].Type -> "Person, Car, ..."
	// items[0].Confidence -> 0.0 (low) -> 1.0 (high)
	// items[0].X -> bounding box
	// items[0].Y -> bounding box
	// items[0].Width -> bounding box
	// items[0].Height -> bounding box
}