An efficient densenet-based deep learning model for Big-4 snake species classification.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Herpetology

Background:

• Snakebite envenomation is a major global health issue, particularly in tropical/subtropical regions, causing millions of cases annually.
• Effective snakebite management and prevention are critical public health priorities.
• The 'BIG FOUR' snake species are responsible for the majority of snakebites in India, highlighting the need for targeted interventions.

Purpose of the Study:

• To develop and evaluate a transfer learning-based image classification algorithm for automatically distinguishing venomous from non-venomous snakes.
• To assess the performance of the DenseNet model using key classification metrics.
• To explore the integration of this AI algorithm into an automated snake-trapping device for enhanced snakebite prevention.

Main Methods:

• Utilized DenseNet architecture for a multiclass image classification task focused on snake identification.
• Employed transfer learning techniques to enhance model performance.
• Evaluated the algorithm's effectiveness using accuracy, F1-score, Recall, and Precision metrics.

Main Results:

• The DenseNet model achieved a notable accuracy rate of 86% for multiclass snake image classification.
• The algorithm demonstrated strong performance across accuracy, F1-score, Recall, and Precision metrics.
• The developed algorithm is suitable for integration into AI-powered snake-trapping systems.

Conclusions:

• The proposed AI algorithm shows significant promise for automated snake identification and capture, offering a novel approach to snakebite prevention.
• This technology has the potential to reduce human-snake conflict and snakebite-related mortality worldwide.
• Further development could lead to practical, non-invasive snakebite mitigation strategies through automated systems.