Visioneer

This application allows you to evaluate the predictions of computer vision models under various conditions.

How it works

1. Drag and drop image from Content Browser Panel.

2. Select model, enable it, tune parameters and assess predictions.

3. Add some augmentations and assess predictions again.

Prerequisites

• C++17 (I use gcc 9.3.0).
• CMake 3.16 or newer is required. Can be installed with apt or build from source.
• Git LFS is required to download ONNX models.
• OpenCV 4.5.3 or newer is required. If you have no OpenCV follow instructions, which are in the next section.
• ONNX Runtime installed from pre-build binaries available here. Instructions are in the next section.
• spdlog, GLFW and ImGui installed via Git Submodules and do not require any configuration.

Installation

# Setup OpenCV
wget -q -O opencv.zip https://github.com/opencv/opencv/archive/refs/tags/4.5.3.zip
unzip -q opencv.zip && rm opencv.zip
cd opencv-4.5.3
mkdir build && cd build
cmake .. -DBUILD_LIST=dnn,imgcodecs -DBUILD_opencv_apps=OFF -DCMAKE_INSTALL_PREFIX=/opt/opencv-4.5.3
sudo cmake --build . --target install -- -j$(nproc)
cd ../..

git clone --recursive https://github.com/teplandr/Visioneer.git
cd Visioneer

# Setup ONNX Runtime
wget -q https://github.com/microsoft/onnxruntime/releases/download/v1.9.0/onnxruntime-linux-x64-1.9.0.tgz
tar xvzf onnxruntime-linux-x64-1.9.0.tgz && rm onnxruntime-linux-x64-1.9.0.tgz
mv onnxruntime-linux-x64-1.9.0 3rdparty/onnxruntime

mkdir build && cd build
cmake ..
cmake --build . -- -j$(nproc)

# Setup ImGui widgets
cp ../resources/internal/imgui.ini .

# Run app
./Visioneer

Dependencies

• spdlog
• glad
• GLFW
• ImGui
• ONNX Runtime
• OpenCV

Acknowledgements

• Yan Chernikov and his Game Engine