Real time object detection with unmanned aerial vehicle

Many studies have been carried out in the field of artificial intelligence and many algorithms have been developed as a result of these studies. Hardware manufacturers are also producing better hardware every day. Thanks to hardware power we have reached today, applications using algorithms that couldn't be run on hardware before are being developed. The large amount of data needed for the realization of supervised learning algorithms in the field of computer vision has become available thanks to mobile phone cameras and the internet. With these data, publicly accessible visual data sets such as PASCAL VOC, ILSVRC, COCO containing images of many classes have been created. However, these datasets do not include the bird's-eye perspective images needed for drone image processing. In this thesis, real-time image processing with unmanned aerial vehicles is aimed. For this purpose, a unique data set consisting of images taken by unmanned aerial vehicles was created. YOLO v4, a convolutional neural network model with real time image processing speed, was chosen and trained with the created dataset. With the developed application, the images taken from the unmanned aerial vehicle were processed with the Yolo v4 Model and real-time object detection was achieved on "person" and "sheep" classes that are included in dataset.