Multi-View Beetle Images for Galerucinae and Alticinae Classification

Multi-view images of 43 chrysomelid beetle species, comprising 23 Galerucinae and 20 Alticinae species, captured from dorsal, lateral, and ventral perspectives. Junhui Liu created this dataset to support the MVBeetle deep learning model for fine-grained taxonomic classification. The dataset was last updated in April 2026.

Use Cases

• Train multi-view CNN models like ResNet18 or VGG16 to classify species into Galerucinae or Alticinae subfamilies using synchronized dorsal, lateral, and ventral images.
• Analyze model attention via Grad-CAM to identify key morphological features, such as jumping legs for Alticinae or antennae for Galerucinae.
• Benchmark classification performance across different network architectures (ResNet50, MobileNetV2) on a dataset of 43 species.
• Study cross-subfamily misclassification rates to assess taxonomic reliability of computer vision models.

Strengths

• Contains synchronized high-resolution images from three distinct anatomical views (dorsal, lateral, ventral) per specimen.
• Covers 43 distinct beetle species, with a balanced representation of 23 Galerucinae and 20 Alticinae species.
• Supports a model achieving a peak classification accuracy of 94.44% ± 0.41% for this taxonomic task.

Limitations

• The dataset is small in scale at 686.9 KB, indicating a limited number of total image samples.
• Species coverage is limited to 43 chrysomelid species, which may not represent broader taxonomic diversity.
• Geographic origin and temporal range of the specimen collection are not specified.

Provenance

Source

Junhui Liu via figshare.

Collection Method

Constructed from synchronized high-resolution images of beetle specimens captured from dorsal, lateral, and ventral views.

Freshness

Last updated in April 2026.