# Comparison of GoogleNet, AlexNet, and ResNet50 Models for Detecting Pyrite Mineral Based on Coal Petrographic Images
> Maharani S.

URL kanonis: https://discover.unhas.ac.id/publications/comparison-of-googlenet-alexnet-and-resnet50-models-for-detecting-pyrite-mineral
Jurnal / Konferensi: Proceedings International Seminar on Intelligent Technology and Its Applications Isitia
Tahun terbit: 2025
DOI: https://doi.org/10.1109/ISITIA66279.2025.11137422
ISSN: 27695492
Citations: 0

## Authors
- Maharani S.

## Abstract
The concentration of pyrite minerals in coal is a contributing factor to the production of acid mine drainage. Acid mine drainage pollution negatively influences water quality in adjacent areas, affecting both the economy and public health in coal mining districts. Consequently, it is essential to perform image processing analyses on coal petrographic pictures to ascertain the existence of pyrite in coal. The image processing model classifies the identification of pyrite minerals from coal petrography discoveries by evaluating three models GoogleNet, AlexNet, and ResNet50 to attain precision in pyrite detection classification. This study comprises a sample of 899 data points obtained from coal petrographic pictures. Analysis of the GoogleNet, AlexNet, and ResNet50 models indicates that the AlexNet model attained the greatest accuracy score of 98.02 % without employing any augmentation techniques. The study results demonstrate that the AlexNet model surpassed other models in the precise identification of pyrite minerals in petrographic pictures. This can assist mining operations in mitigating water pollution resulting from acid mine drainage. This can assist mining operations in mitigating water pollution resulting from acid mine drainage.

## Keywords
- Petrography
- Pyrite
- Mineral
- Geology
- Coal
- Geochemistry
- Mineralogy
- Computer science
- Artificial intelligence
- Materials science
- Archaeology
- Metallurgy
- Geography

---
Sumber: Discover Unhas — RIMS Universitas Hasanuddin.
Saat mengutip, gunakan DOI bila tersedia atau URL kanonis di atas.