Efficient Drones-Birds Classification Using Transfer Learning
Taha Mahdy Mohamed, Ibraheem Mubarak Alharbi · 2023
Transfer learning is one of the recent deep learning technologies that can saves most of the time consumed in training deep networks. It could be performed by using pre-trained deep networks. In this paper, we will propose a new model for drones-bird classification based on transfer learning. We modify and evaluate three popular pre-trained deep models using dataset consists of drones and birds. The performance of each pre-trained network is evaluated and compared. Results show that, these pretrained networks could be adopted efficiently in classification of the addressed dataset. Additionally, the accuracy of the ResNet18 outperforms other evaluated networks. The accuracy and F-Score of ResNet18 exceeds 98% in all cases. Also, the other models have an excellent performance. There are many important applications of the problem addressed especially in security, defense, and surveillance. So, accurate classification of drones and birds is very important.