Reptile Identification for Endemic and Invasive Alien Species Using Transfer Learning Approaches.

Area of Science:

• Ecology
• Computer Science
• Biodiversity Conservation

Background:

• The Canary Islands are a biodiversity hotspot with numerous endemic reptile species.
• Invasive alien reptile species pose a significant threat to the archipelago's unique ecosystems.
• Current control methods rely on sporadic sightings, hindering effective management of invasive species.

Purpose of the Study:

• To develop an automated system for identifying endemic and invasive reptile species in the Canary Islands.
• To explore the application of deep learning (DL) techniques for reptile species recognition.
• To address the lack of automated identification tools for Canary Islands reptiles.

Main Methods:

• Implementation of various neural network models utilizing transfer learning approaches.
• Focus on deep learning (DL) techniques for automatic species recognition.
• Evaluation of different models for their performance in reptile identification.

Main Results:

• The study successfully implemented and compared multiple neural network models.
• The EfficientNetV2B3 base model demonstrated superior performance in species identification.
• The best model achieved a mean Accuracy of 98.75% in identifying target reptile species.

Conclusions:

• Deep learning offers a promising solution for automated reptile species identification in the Canary Islands.
• The EfficientNetV2B3 model shows high potential for practical application in conservation efforts.
• Automated detection systems can significantly improve the management of invasive alien species and protect endemic biodiversity.