Multi-Class Parrot Image Classification Including Subspecies with Similar Appearance.

Area of Science:

• Conservation Biology
• Computer Science
• Artificial Intelligence

Background:

• Endangered species populations are declining due to climate change and human activities.
• The Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) aims to protect threatened species.
• Deep learning-based image recognition offers potential for wildlife conservation.

Purpose of the Study:

• To develop an automated image classification method for 11 endangered parrot species.
• To address challenges in distinguishing visually similar subspecies.

Main Methods:

• Collected and curated a dataset of endangered parrot images from the internet and a zoo.
• Utilized a deep learning approach with a dataset split for training, validation, and testing.
• Applied data augmentation to enhance the dataset and prevent overfitting.
• Compared various Convolutional Neural Network (CNN) architectures (VGGNet, ResNet, DenseNet) using the Single Shot Detector (SSD) model.

Main Results:

• The DenseNet18 architecture demonstrated the best performance among the tested models.
• Achieved a mean Average Precision (mAP) of approximately 96.6% on the test set.
• The model exhibited an efficient inference time of 0.38 seconds.

Conclusions:

• The proposed automated image classification method, particularly using DenseNet18, is highly effective for identifying endangered parrot species.
• This technology can significantly support conservation initiatives for CITES-listed birds.
• The approach offers a scalable solution for monitoring and protecting endangered wildlife through advanced AI.