With the advancement of technology, drones/unmanned aerial vehicles (UAVs) are widely used in various fields. Uncontrolled flights of UAVs pose serious security issues due to their potential use in illegal activities. Therefore, drone monitoring and automatic detection are crucial. In this study, a YOLOv5-based Cross convolution with BottleneckCSP (CB-YOLOv5) model is proposed for drone detection. The Bottleneck CSP module is employed for higher accuracy in multi-scale targets, while the cross convolution module is used to reduce the number of parameters and increase the inference speed of the model. The images are preprocessed with auto-orientation and resize. Then, the processed images are fed into CB-YOLOv5 to extract their deep features. Meanwhile, the proposed method is compared with state-of-the-art techniques to validate its effectiveness. The proposed method improves the mAP performance of the drone detection model by at least 5.08%.