MLDAAPP: machine learning data acquisition for assessing population phenotypes.

Area of Science:

• Genetics and Genomics
• Computational Biology
• Animal Behavior

Background:

• Manual collection of organismal phenotypic data limits research scale, reproducibility, and introduces bias.
• Existing computer vision tools lack flexibility and scalability for diverse phenotypic data acquisition.

Purpose of the Study:

• To introduce MLDAAPP, a novel software package for automated phenotypic data collection using computer vision.
• To provide a flexible and scalable solution for generating diverse phenotypic data from populations.

Main Methods:

• Development of MLDAAPP, a tool suite built upon the YOLOv8 object detection framework.
• Application of MLDAAPP for collecting quantitative data such as counts and movement metrics from various organisms.

Main Results:

• MLDAAPP demonstrates accuracy in collecting phenotypic data, even from challenging image and video qualities.
• The tool effectively handles data from both laboratory and field environments.
• Successful application in generating data for phenotypes like Drosophila fecundity and animal activity.

Conclusions:

• MLDAAPP significantly improves the reproducibility and scale of phenotypic data generation in biological research.
• This tool addresses key limitations in current data acquisition methods, enabling broader and more in-depth population studies.