[Paper Review] A Step Towards Worldwide Biodiversity Assessment: The BIOSCAN-1M Insect Dataset

Zahra Gharaee, ZeMing Gong+14 more|arXiv (Cornell University)|Jul 19, 2023

Species Distribution and Climate Change9 citations

TL;DR

The paper introduces the BIOSCAN-Insect Dataset, a curated million-image collection with expert taxonomic labels and associated genetic barcodes, and presents a baseline classifier analysis for image-based taxonomic assessment.

ABSTRACT

In an effort to catalog insect biodiversity, we propose a new large dataset of hand-labelled insect images, the BIOSCAN-Insect Dataset. Each record is taxonomically classified by an expert, and also has associated genetic information including raw nucleotide barcode sequences and assigned barcode index numbers, which are genetically-based proxies for species classification. This paper presents a curated million-image dataset, primarily to train computer-vision models capable of providing image-based taxonomic assessment, however, the dataset also presents compelling characteristics, the study of which would be of interest to the broader machine learning community. Driven by the biological nature inherent to the dataset, a characteristic long-tailed class-imbalance distribution is exhibited. Furthermore, taxonomic labelling is a hierarchical classification scheme, presenting a highly fine-grained classification problem at lower levels. Beyond spurring interest in biodiversity research within the machine learning community, progress on creating an image-based taxonomic classifier will also further the ultimate goal of all BIOSCAN research: to lay the foundation for a comprehensive survey of global biodiversity. This paper introduces the dataset and explores the classification task through the implementation and analysis of a baseline classifier.

Motivation & Objective

• Motivate and enable worldwide biodiversity assessment through a large, hand-labeled insect image dataset.
• Provide taxonomic labels that are hierarchical and fine-grained to reflect real-world biodiversity complexity.
• Incorporate genetic information (raw nucleotide barcodes and barcode index numbers) as auxiliary data for taxonomy tasks.
• Demonstrate the feasibility of image-based taxonomic classification to support global biodiversity surveys.

Proposed method

• Assemble and curate a million-image insect dataset with expert taxonomic classification.
• Attach associated genetic information including raw nucleotide barcode sequences and barcode index numbers to each record.
• Highlight long-tailed class-imbalance characteristics inherent to biological datasets.
• Characterize the hierarchical, fine-grained taxonomy challenge for classification tasks.
• Implement and analyze a baseline image-based taxonomic classifier on the dataset.

Experimental results

Research questions

• RQ1 Can image-based methods achieve taxonomic classification across a hierarchical, fine-grained insect taxonomy?
• RQ2 How does long-tailed class distribution affect image-based biodiversity classification performance?
• RQ3 What is the role of accompanying genetic barcode information in supporting image-based taxonomy?
• RQ4 Is a baseline classifier capable of providing a useful starting point for global biodiversity surveys using BIOSCAN data?

Key findings

• The BIOSCAN-Insect Dataset is presented as a curated million-image resource for taxonomic assessment.
• The dataset exhibits long-tailed class-imbalance characteristics typical of biological data.
• Taxonomic labeling follows a hierarchical, highly fine-grained scheme at lower levels.
• A baseline image-based classifier is implemented and analyzed on the dataset to establish a starting point for future work.
• The study emphasizes the dataset's potential to support broader efforts toward worldwide biodiversity surveys.