EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Performance of Two Beginner-Accessible AI Image Classifiers for Marine-Debris Identification on Beaches of the Texas Gulf Coast

Sophie Qiao

American Journal of Capstone Research · 2026

DOI: 10.70251/1382 ↗

Kurz gefasst

Fachartikel in der Zeitschrift American Journal of Capstone Research, 2026. Verfasst von Sophie Qiao. Die Arbeit nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Marine debris poses persistent environmental and human-health risks, while recent advances in artificial intelligence (AI) have expanded the accessibility of image-classification tools. This study investigated the extent to which two beginner-accessible AI image classifiers, trained by a high school student, could distinguish between oceanic debris and natural marine objects on beaches along the Gulf of Mexico coast in southern Texas. Google Teachable Machine and Roboflow were each trained on 500 internet-sourced images (250 debris, 250 natural marine objects) and evaluated on the same 125 locally collected test images (59 debris, 66 natural marine objects). Teachable Machine correctly classified 116 of 125 images (92.8%, 95% CI [86.9%, 96.2%]), and Roboflow correctly classified 119 of 125 images (95.2%, 95% CI [89.9%, 97.8%]); both exceeded the majority-class chance baseline of 52.8% (one-tailed binomial tests, p < 0.001 for both). A paired McNemar’s exact test found no statistically significant difference between the two classifiers (p = 0.61), despite Roboflow’s numerically higher accuracy, sensitivity, and F1 score. These findings suggest that beginner-accessible AI tools can support accurate binary image classification of marine debris in a localized context, though the limited and non-independently verified training and test datasets, single-rater ground truth, single geographic region, and short collection window constrain generalizability. Future research should test larger, independently validated datasets across multiple locations and seasons before drawing conclusions about real-world environmental-monitoring applications.

Zitieren

APA

Qiao, S. (2026). Performance of Two Beginner-Accessible AI Image Classifiers for Marine-Debris Identification on Beaches of the Texas Gulf Coast. American Journal of Capstone Research. https://doi.org/10.70251/1382

Vancouver

Qiao S. Performance of Two Beginner-Accessible AI Image Classifiers for Marine-Debris Identification on Beaches of the Texas Gulf Coast. American Journal of Capstone Research. 2026. doi:10.70251/1382

BibTeX
@article{Qiao2026,
  author = {Qiao, Sophie},
  title = {Performance of Two Beginner-Accessible AI Image Classifiers for Marine-Debris Identification on Beaches of the Texas Gulf Coast},
  journal = {American Journal of Capstone Research},
  year = {2026},
  doi = {10.70251/1382}
}
Kostenlos starten →