Image-based recognition of parasitoid wasps using advanced neural networks

Area of Science:

• Entomology
• Biodiversity Research
• Computational Biology

Background:

• Hymenoptera insects exhibit high diversity, yet ~80% of species remain undescribed.
• Traditional morphological taxonomy is slow, hindering biodiversity assessment.
• DNA barcoding has improved insect identification, but faster methods are needed.

Purpose of the Study:

• To develop and validate an image-based automated system for identifying Diapriidae wasps.
• To leverage DNA barcoding for ground truth data to train machine learning models.
• To assess the potential of artificial intelligence in accelerating insect taxonomy.

Main Methods:

• Specimens of Diapriidae wasps were photographed and identified using DNA barcoding.
• A neural network was trained using this image and taxonomic data.
• Three different neural network architectures were evaluated and optimized for classification.

Main Results:

• The system achieved 96% average accuracy in classifying 11 genera of Diapriidae and other Hymenoptera.
• Automated sex classification of specimens reached an accuracy exceeding 97%.
• The proof-of-concept demonstrated the efficacy of image recognition for wasp identification.

Conclusions:

• Automated image recognition offers a rapid and accurate method for insect identification, particularly for small species like Diapriidae.
• This technology can significantly accelerate taxonomic research and biodiversity monitoring.
• AI-driven approaches hold great promise for addressing the taxonomic impediment in entomology.