bwh-detector

A #MachineLearning project with the aim to mark and count #Bees, #Wasps and #Hornets (appriviation used : #BWH) in a picture.

This Project has the following dependency: https://github.com/AlexeyAB/darknet or the original https://pjreddie.com/darknet/yolo/

Darknet is a tool which implements a neural-net; The core of every detection-set is

• a Weights-File
• the corresponding config-File
• the correspondig Class-Naming-File

The training of the weights-file is time- and power-consuming; and espacialy for Bees, Wasps and Hornet (#bwh) not an easy thing to do;

Project-Content

Config and Weights

This Repository contains several config and weights-Files in ongoing state. The quality is actually good for the use it was initialy created:

Intended Use of this repositoy

The BeeCam is a 4MP-Webcam. A lower resulotion may work but quality may differ; The distance to the #Bee-Hive is about 50-60 cm.

Problems

Resolution of the Object

The taken picture can have variete of sizes;

Object-Distance

Depending of the picture it can contain thounsands of bwh (where every #bwh is only some pixels big) - or just one big.

Wasps

Wasps and Bees are a Category - and there are many different kinds; This weights-file are trained on "German Wasps / Vespula germanica" and "Gemeine Wespe / Vespula vulgaris"

DATASET

The main issue - i did have - was a big enough dataset espacially for #Wasps and #Hornets. I did uses the following tool to #Label all Images: https://github.com/Cartucho/OpenLabeling The labeled image-dataset can be requested by issue;

FAQ

This project assumes that the "darknet-binary" is in the $PATH-Variable of your OS.

Windows-Installation of Alexeys darknet-fork

On Windows the most easy way to build it - is this; But it may depend on updates of the darknet-repository.

• Download Visual-Studio 2017 and install;
• If not available -- install GIT
• optional

• Download CUDA and install;
• Download CUDNN and unzip it into %CUDA_PATH%\..\;

• Clone VCPKG
• set environment-variable VCPKG_DEFAULT_TRIPLET : x64-windows
• set environment-variable VCPKG_ROOT : path-where-you-put-vcpkg
• build VCPKG like described in quickstart.
• Open a Terminal or Powershell enter cd %VCPKG_ROOT% and run ./vcpkg install darknet[opencv] or ./vcpkg install darknet[opencv-cuda]

Be patient -- it did take about 2-3 hours to compile everything

• Add to environment variable PATH the following: %VCPKG_ROOT%\installed\%VCPKG_DEFAULT_TRIPLET%\bin;%VCPKG_ROOT%\installed\%VCPKG_DEFAULT_TRIPLET%\tools\darknet

Now you should be able to start darknet.exe from every fresh CMD-Term or Powershell.