A Comparative Evaluation of ResNet-101 and Inception v3 Architectures for Wildlife Object Detection
IOSR Journal of Computer Engineering · 2026-03 · 28 (3) · 35-44
DOI: 10.9790/0661-2803023544 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-03. Verfasst von Malach Obisa und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Wildlife detection and identification are critical components in biodiversity conservation and ecological monitoring. This study evaluates the effectiveness of two deep learning architectures ResNet-101 and Inception v3 in classifying and detecting wildlife species under diverse environmental conditions. Both models were trained on a curated wildlife dataset with standardized preprocessing techniques, including resizing images to a maximum of 800 pixels, RGB format conversion, and transformation into PyTorch tensors. An 80:20 training-validation split was applied. The ResNet-101 model achieved a classification accuracy of 93.5% and a mean Average Precision (mAP) of 0.91, while the Inception v3 model attained a slightly higher accuracy of 94.7% and a mAP of 0.916. Evaluation metrics such as confusion matrices, precision-recall curves, and class-wise performance indicate that both models are suitable for wildlife object detection, with Inception v3 showing a marginal edge. This work contributes benchmark results for future wildlife detection models and provides insights into the tradeoffs between deep feature extraction and multi-scale processing in convolutional neural networks.
Zitieren
APA
Obisa, M., Osero, D. B., & Too, D. E. (2026). A Comparative Evaluation of ResNet-101 and Inception v3 Architectures for Wildlife Object Detection. IOSR Journal of Computer Engineering, 28(3), 35-44. https://doi.org/10.9790/0661-2803023544
Vancouver
Obisa M, Osero DB, Too DE. A Comparative Evaluation of ResNet-101 and Inception v3 Architectures for Wildlife Object Detection. IOSR Journal of Computer Engineering. 2026;28(3):35-44. doi:10.9790/0661-2803023544
BibTeX
@article{Obisa2026,
author = {Obisa, Malach and Osero, Dr. Benard and Too, Dr. Edna},
title = {A Comparative Evaluation of ResNet-101 and Inception v3 Architectures for Wildlife Object Detection},
journal = {IOSR Journal of Computer Engineering},
year = {2026},
volume = {28},
number = {3},
pages = {35--44},
doi = {10.9790/0661-2803023544}
}
EUVIMED