Improving Wild Pig Detection Through Data Augmentation and Thermal Imagery: A Comparative Study of Model Performance.
Abstract
This study presents a comparative analysis of wild pig detection using a multimodal fusion framework, integrating YOLOv8 with both RGB and thermal imagery. We investigate the impact of data augmentation techniques and GroundDINO-based auto-labeling on detection accuracy, aiming to enhance the model's robustness in diverse environmental conditions. Our results demonstrate significant improvements in precision and recall when incorporating thermal imagery and augmentation, especially in challenging environmental scenarios, such as low light and dense vegetation.
Cite
Text
Gonzalez and Kim. "Improving Wild Pig Detection Through Data Augmentation and Thermal Imagery: A Comparative Study of Model Performance.." NeurIPS 2024 Workshops: LXAI, 2024.Markdown
[Gonzalez and Kim. "Improving Wild Pig Detection Through Data Augmentation and Thermal Imagery: A Comparative Study of Model Performance.." NeurIPS 2024 Workshops: LXAI, 2024.](https://mlanthology.org/neuripsw/2024/gonzalez2024neuripsw-improving/)BibTeX
@inproceedings{gonzalez2024neuripsw-improving,
title = {{Improving Wild Pig Detection Through Data Augmentation and Thermal Imagery: A Comparative Study of Model Performance.}},
author = {Gonzalez, Alhim Adonai Vera and Kim, Donghoon},
booktitle = {NeurIPS 2024 Workshops: LXAI},
year = {2024},
url = {https://mlanthology.org/neuripsw/2024/gonzalez2024neuripsw-improving/}
}