Saved in:
Bibliographic Details
Main Authors: Gardiner, Ross J, Mougeot, Guillaume, Rowlands, Sareh, Simmons, Benno I, Helsing, Flemming, Høye, Toke Thomas
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.20089
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914009562742784
author Gardiner, Ross J
Mougeot, Guillaume
Rowlands, Sareh
Simmons, Benno I
Helsing, Flemming
Høye, Toke Thomas
author_facet Gardiner, Ross J
Mougeot, Guillaume
Rowlands, Sareh
Simmons, Benno I
Helsing, Flemming
Høye, Toke Thomas
contents Labelling images of Lepidoptera (moths) from automated camera systems is vital for understanding insect declines. However, accurate species identification is challenging due to domain shifts between curated images and noisy field imagery. We propose a lightweight classification approach, combining limited expert-labelled field data with knowledge distillation from the high-performance BioCLIP2 foundation model into a ConvNeXt-tiny architecture. Experiments on 101 Danish moth species from AMI camera systems demonstrate that BioCLIP2 substantially outperforms other methods and that our distilled lightweight model achieves comparable accuracy with significantly reduced computational cost. These insights offer practical guidelines for the development of efficient insect monitoring systems and bridging domain gaps for fine-grained classification.
format Preprint
id arxiv_https___arxiv_org_abs_2508_20089
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Bridging Domain Gaps for Fine-Grained Moth Classification Through Expert-Informed Adaptation and Foundation Model Priors
Gardiner, Ross J
Mougeot, Guillaume
Rowlands, Sareh
Simmons, Benno I
Helsing, Flemming
Høye, Toke Thomas
Computer Vision and Pattern Recognition
Labelling images of Lepidoptera (moths) from automated camera systems is vital for understanding insect declines. However, accurate species identification is challenging due to domain shifts between curated images and noisy field imagery. We propose a lightweight classification approach, combining limited expert-labelled field data with knowledge distillation from the high-performance BioCLIP2 foundation model into a ConvNeXt-tiny architecture. Experiments on 101 Danish moth species from AMI camera systems demonstrate that BioCLIP2 substantially outperforms other methods and that our distilled lightweight model achieves comparable accuracy with significantly reduced computational cost. These insights offer practical guidelines for the development of efficient insect monitoring systems and bridging domain gaps for fine-grained classification.
title Bridging Domain Gaps for Fine-Grained Moth Classification Through Expert-Informed Adaptation and Foundation Model Priors
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2508.20089