Abstract:
In the realm of avian conservation, this thesis embarks on a pioneering journey to enhance
the classification of pigeon species within Bangladesh. Leveraging the powerful Xception
model, we present a breakthrough approach that attains an exceptional testing accuracy of
99.47% and minimal loss of 0.025. Our study encompasses a comprehensive dataset of
7500 images, spanning 15 pigeon species, and employs transfer learning for swift and
reliable classification. While the results underscore the efficacy of our approach, the study
acknowledges the challenge of subjective criteria in species classification and calls for
future exploration into enhancing interpretability. Ethical considerations are central to our
findings, advocating transparent communication with conservationists and the
establishment of stringent ethical guidelines for responsible technology application in
avian conservation. This research, a significant stride at the intersection of technology and
ethics, not only contributes to avian conservation but also lays the groundwork for future
investigations, paving the way for a sustainable future in avian species management and
urban biodiversity preservation.
