Bangladesh is a land with a wide range of ecosystems, the country has a variety of topographic features such as mountains, lakes, evergreen and deciduous forests, and swamp forests. The country is also gifted with a coastline that spans the southern region of the country. This diverse land is the home of some of the most beautiful birds. Our main goal is to propose a convolutional neural network-based approach to identifying native birds of Bangladesh. Identifying native birds is crucial for biodiversity conservation, as it helps monitor species populations, track ecosystem health, and detect environmental changes. In ecological studies, native bird data offer insights into habitat quality and inform conservation strategies to protect both the birds and their ecosystems. To achieve our desired goal, we have done the entire research work using three different CNN architectures and they are VGG16, MobileNetV2, and ResNet50. The dataset used is the collection of images of 18 native Bangladeshi Bird species. Augmentation has been performed to deal with the insufficiency of image data. The models have been trained with 2152 image data of 18 different native Bangladeshi bird species and tested with 539 image data of 18 different native Bangladeshi bird species. Among all the used models, ResNet50 performs the best with a test accuracy of 98%. The entire work shows that the working method is optimal and efficient for recognizing native birds of Bangladesh.