User-Interface Unit

The goal of this project is to develop a user-friendly tool that automates the placement of anatomical landmarks on lizards. Landmarks are annotated points on a specimen, such as joints, that enable researchers to measure and compare individuals. These comparisons support the identification of broader morphological patterns within and across ecosystems. The tool integrates a pose-estimation model to predict landmark locations and is connected to an interactive user interface that allows users to edit, visualize, and download the model’s predictions. By streamlining the data collection process, this tool helps researchers work more efficiently, allowing them to focus on analysis and discovery rather than manual annotation.

To get the most recent updates on this project, visit our Wikipedia page: https://humanaugmentedanalyticsgroup.miraheze.org/wiki/Stroud_Lab_Projects/Lizard_Morph

The goal of this project is to develop an automated machine learning pipeline that applies morphometric measurements to lizard toepad scans, building directly on the framework used for the lizard x-ray project. We currently perform this process manually, but now have a large, annotated dataset available for researchers to use in developing and refining an automated pipeline. The same pipeline architecture and graphical user interface (GUI) from the lizard x-ray project will be leveraged for consistency and efficiency. At present, a YOLO model has already been trained to detect and localize the toepads in images. Students will have the opportunity to iterate on and improve this model to increase accuracy and robustness. The next step in the pipeline will involve building a model capable of automatically placing thin-plate spline (TPS) landmark points on the detected toepads, enabling high-throughput morphometric analysis.