Effectiveness of Inception V3and Mo bileN et V2Models in Classifying Bird Species Based on Physical Characteristics

Birds are wildlife that are easily found in every environment. With bird populations growing in abundance but threatened by extinction, bird species identification is crucial for nature conservation efforts. Proper classification can provide a strong basis for protecting habitats and implementing effective conservation policies. This study uses a dataset consisting of 525 bird species with a total of 84,635 training images, 2625 test images, and 2625 validation images. Two Convolutional Neural Network (CNN) architectures, namely Inception V3 and MobileN et V2, were compared to evaluate the effectiveness of using the model in classifying bird species. Previous studies have shown that the use of CNN s, especially the Inception Resnet V2 architecture, has resulted in high accuracy rates in bird species recognition. The results of this study show that MobileN et V2 achieved higher accuracy (94.93 %) compared to Inception V3 (91.16
^%
) in classifying bird species on the test dataset. The loss value of MobileNet V2 also tends to be lower, indicating better performance in training. These results can indicate that the Mobilenet V2 architecture model has an architecture or configuration that is more suitable for the characteristics of the test dataset used, resulting in better results in terms of accuracy and loss application of machine learning technology, especially CNN, provides new possibilities for detecting and identifying bird species efficiently. This research is expected to be a reference regarding accurate identification results, which can be used as a basis for understanding the population, habitat, and behavior of bird ecosystems. This research, which uses the Inception V3 and MobileN et V2 architecture models, is expected to contribute to the development of more efficient and accurate bird species identification methods in supporting nature conservation and wildlife protection efforts.