The rapid development of unmanned aerial vehicles (UAVs) or drone technology, which can move at different speeds and altitudes and are more affordable to civilians, has resulted in a rapid spread of their civilian use. These aircraft, due to the materials they carry on board, can be used to cause negative effects, posing a threat to the state’s infrastructure and industries and the security of the civilian population. For this reason, automated detection of drones is considered an essential task within the framework of air security systems. Because of the great similarity between drones and birds, both physically and behaviorally, and they are often confused, this study proposes a new method based on visual range and recent developments in deep neural networks (DNNs) to effectively detect and identify drones and birds. The proposed approach was evaluated with image data and found to be a better detector compared to previous detection systems. Four models were trained and tested (Resnet50, Resnet18, Mobilenetv2, and alexnet) and their results were compared in detecting drones and differentiating them from birds. To train the network, a diverse and comprehensive dataset (BVD) was used, which includes a wide range of scenarios, under different environmental conditions, multiple drone types with an equal distribution of drone and bird instances, and includes label information. The suggested deep learning method can detect a drone and distinguish it from a bird with an accuracy of 95% (Resnet50), 91% (Resnet18), 90% (Mobilenetv2), and 85% (alexnet). The values of average precision were also reported as 95% (Resnet50), 95% (Resnet18), 89% (Mobilenetv2), and 82% (alexnet).