The project Testbench for UAVs is an opensource project written as a part of Research and Development (RnD) project for Autonomous Systems (MSc) studies at Hochschule Bonn-Rhein-Sieg (H-BRS). This project presents a testbench for UAVS. It can be used to test different datasets with different algorithms. Currently it supports 2 datasets and 1 algorithm but it can be easily extended by implementing the interfaces it provides.

• Testbench for UAVs
  • About
  • Table of Contents
  • Installation
    • Setup the Datasets
    • Setup the Algorithms
    • Run the Testbench
  • Usage

To install the project first clone the repository. And move into it.

git clone [email protected]:ssmabidi/Interface.git
cd Interface

Next create a folder Datasets in the Interface folder and download the datasets. We need to download The Zurich Urban Micro Aerial Vehicle Dataset and The AU-AIR Dataset

mkdir Datasets
cd Datasets

Zurich Dataset

To download and setup the zurich dataset run the following commands

wget "web.eee.sztaki.hu/~majdik/AGZ/AGZ.zip"
unzip AGZ.zip

After the unzip is complete the Zurich Dataset is ready for use in the Testbench.

Au-Air Dataset

To download and setup Au-Air dataset first create a auAir folder

mkdir auAir
cd auAir

Then download the AU-AIR images from the following link and extract them into this folder.

Images: https://drive.google.com/open?id=1pJ3xfKtHiTdysX5G3dxqKTdGESOBYCxJ (2.2 GB)

Extract this zip file in auAir directory. Following commands can be used for extraction.

cd auAir
unzip auair2019data.zip 

After the download and unzip is complete, then download the AU-AIR annotations and place them in auAir folder.

Annotations (V.1.1): https://drive.google.com/file/d/1GyoBK-NalDFfAtRt9LO6FBujbObyaZLv/view?usp=sharing (55 MB)

Make sure the name of annotations file is annotations_v1.1.json

The project currently supports just one Algorithm YOLOv4: Optimal Speed and Accuracy of Object Detection.

YOLOv4

The project uses a yolo library which is build with GPU enabled. Make sure CUDA is installed in order for algorithm part to run successfully. Then download pre trained YOLOv4 weights using the following commands.

cd ../../yoloFiles/
wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights

Install the requirements present in requirements.txt using pip

cd ..
pip3 install -r requirements.txt

To run the application navigate to the root folder and start the project using following commands

python3 webPage/index.py

Then open any web browser and navigate to http://127.0.0.1:8050/