Bird detection is a crucial problem encountered in many applications. Object detection methods based on machine learning have shown promising performance on this task. However, the number and types of bird classes are very important and can vary significantly across different applications. Moreover, overfitting is a common issue in machine learning that occurs when a model fits too closely the training data, resulting in poor generalization to new, unseen data. To overcome this problem, we propose a new bird dataset that contains a diverse number and types of bird species that are well-suited for different applications. This helps the model learn the underlying patterns in the data rather than memorizing the training samples. The YOLOv5s model was trained on our proposed dataset. The obtained results of bird detection achieved a precision of up to 100% for some bird species, demonstrating the effectiveness of the proposed database.