Transfer Learning for Bird Species Identification

Abstract

Monitoring and conservation of bird species play a crucial role in preserving biodiversity and maintaining the balance of the ecosystem. To address this, we have developed an automatic bird recognition system, known as the birdhouse, using the Arduino and Keras deep learning frameworks. The system is equipped with a PIR sensor that activates an ESP-32 camera to capture an image of the bird and send it to the server for processing. The deep learning model, trained using transfer learning with the MobileNetV2 architecture, is deployed with the python flask framework and is able to accurately predict the bird species with 95% test accuracy. The identified bird species is then notified to the users via the telegram application, along with the captured image of the bird. MobileNetV2 is a powerful deep learning architecture that is well-suited for deployment on resource-constrained devices such as the ESP-32 camera used in the birdhouse system. The use of transfer learning allows the model to be trained on a large dataset and then fine-tuned for the specific task of bird species recognition.