Deep learning based real time identification system

Nowadays, object detection and tracking has become one of the most studied areas. The reason for this is that it is of critical importance in the use of security, defense, medical, robotic and autopilot vehicles encountered in daily life. For this purpose, many decision support systems or expert systems using artificial intelligence and machine learning have been tried to be developed. Recently, depending on the developments in the field of deep learning and hardware, many effective and reliable object detection and tracking systems have been developed. In order to increase the performance of deep learning algorithms and to develop more economical systems, studies in the field of object detection and tracking continue rapidly. Scientists who design new algorithms are working to develop models that will reduce the computational cost and increase performance. Developed new models can be used in real-time applications. The aim of this thesis is to increase the performance of online object tracking systems by using deep learning approaches and to develop an application that can be used in daily life. The thesis consists of two parts. First, a comparison of existing optimization algorithms on the multi-object tracking deep learning model FairMOT is made. Optimization algorithms, which are among the hyper-parameters, have been tried and a more successful model has been obtained. Different optimization methods were applied to the FairMOT deep learning model using the MOT20 dataset. The best results were obtained using the RMSprop optimization algorithm. In the second part, a deep learning based real-time face tracking system working on multiple camera systems has been developed. SCRFD model was used for face detection and ArcFace model was used for face detection. For object tracking, the DeepSORT algorithm was used by making use of face detection and face recognition models. Apache Kafka stream processing system and Socket.IO bidirectional communication library were used to process data in real time. As a result of the study, a successful face detection, recognition and tracking system application has been developed.