Red Abalone Ultrasound Images for Sex Classification via Deep Learning

Edwin A. Solares presents a dataset of 246 high-quality ultrasound images from 44 red abalone (Haliotis rufescens) individuals, used to benchmark machine learning models for non-invasive sex determination. The dataset was used to evaluate seven convolutional neural network architectures, with YOLOv8 achieving the highest test accuracy of 85.7%. The dataset was last updated on 2026-05-18.

Use Cases

• Benchmarking deep learning models for biological image classification based on ultrasound images.
• Developing non-invasive sex determination tools for aquaculture breeding programs based on gonadal tissue features.
• Training edge-deployable models for real-time field classification based on inference on NVIDIA Jetson devices.
• Studying model performance on speckle noise inherent in ultrasound imaging based on the described custom architectures.

Strengths

• Contains 246 labeled ultrasound images from 44 individual abalone.
• Data partitioning was performed by individual identity to prevent artificially inflated accuracy from image leakage.
• Benchmarks seven specific CNN architectures (VGG16, VGG19, ResNet50, ResNet101, YOLOv8, YOLOv11, and a custom CNN).
• Achieved a test accuracy of 85.7% with YOLOv8, with reported precision and recall metrics for male and female classes.

Limitations

• Row count and column-level documentation are absent; field semantics must be inferred after download.
• The dataset is small in scale at 2.2 MB, which may limit the complexity of models that can be trained.
• Freshness should be verified as the last update date is in the future (2026-05-18).

Provenance

Source

Edwin A. Solares via figshare

Collection Method

Ultrasound imaging of red abalone (Haliotis rufescens) individuals.

Freshness

Last updated 2026-05-18 05:35:52