Towards scalable aerial detection: Augmenting datasets for real-time animal health monitoring

Authors: PILLAI, NISHA - Mississippi State University; SMITH, HARRISON - Mississippi State University; Ashworth, Amanda; Gowda, Prasanna; Owens, Phillip; Rivers, Adam; NADURI, BINDU - Mississippi State University; RAMKUMAR, MAHALINGAM - Mississippi State University

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 12/4/2024
Publication Date: 12/5/2024
Citation: Pillai, N., Smith, H., Ashworth, A.J., Gowda, P.H., Owens, P.R., Rivers, A.R., Naduri, B., Ramkumar, M. 2024. Towards scalable aerial detection: Augmenting datasets for real-time animal health monitoring. Abstract. Plant & Animal Genome Conference, San Diego, California.

Interpretive Summary:

Technical Abstract: Animal health monitoring research is severely limited by scarce datasets and expensive data collection. We propose a novel approach using transformer-based instance segmentation and advanced data augmentation techniques to synthetically expand animal health monitoring datasets. Our computer vision methodology generates authentic artificial training data that substantially reduces data acquisition costs. Experimental evaluation validate the effectiveness of our artificially generated datasets in animal detection applications, that demonstrate performance metrics closely approximating ground truth data. This research provides a scalable framework for dataset generation, potentially transforming animal health monitoring through a cost-effective and reproducible solution.