Automated Estimation of Snail Shell Volume and Morphology

This project develops an Artificial Intelligence (AI)-based pipeline for automated shell volume estimation and morphological feature extraction to support biological research on gastropod and crustacean specimens. Thousands of shell images have been collected using an Olympus TG-6 camera, originally intended to capture top, bottom, and side views of each specimen. However, practical constraints during field imaging have resulted in challenges such as repeated images, inconsistent viewing angles, and missing third-view data. These limitations require a robust AI approach capable of extracting meaningful three-dimensional measurements even when image inputs are imperfect.

The proposed workflow uses image segmentation techniques, including models such as the Segment Anything Model (SAM), to isolate shell silhouettes from available views. By analysing segmented front and side profiles, computational methods are applied to reconstruct the specimen’s approximate 3D geometry through layer-wise rotational or incremental integration. Preliminary experiments using this pipeline have produced plausible volume estimates, demonstrating strong potential for automated analysis.
The project will further refine the pipeline to improve segmentation robustness, enhance shape reconstruction accuracy, and extract additional morphological traits relevant to ecological and evolutionary studies. These include surface area, growth band patterns, colour descriptors, and indicators of predation repair.
The ultimate goal is to develop a scalable and reliable system capable of analysing large image collections to support research on shell biology, trait variation, and 3D exoskeleton reconstruction. This work highlights how AI and computer vision can accelerate biological data analysis and enable new insights into specimen morphology.

Project Team and Collaborators:

Sue Ann Watson, Quanwei Liu, Tao (Kevin) Huang