Deep Sporenet: A Lightweight Few-Shot CNN For Illumination-Robust Fungal Species Identification on Mobile Devices

Fungal infections in plants and humans pose major challenges to food security and clinical diagnostics, yet their identification still depends on expert microscopy and culture-based methods that are slow and equipment-intensive. Existing deep learning models for fungal classification achieve high accuracy under controlled imaging conditions but fail under illumination shifts, class imbalance, and low-data regimes typical of field or point-of-care scenarios. This study introduces Deep SporeNet, a compact convolutional neural network (CNN) with a few-shot episodic learning head and illumination-robust preprocessing, designed for mobile deployment in agricultural and clinical environments.
The proposed framework integrates (i) color constancy and stain normalization to counter variable lighting, (ii) a MobileNetV2/EfficientNet-Lite backbone for efficient feature extraction, (iii) a Prototypical Network head for low-sample fungal taxa recognition, (iv) entropy-based test-time adaptation (TENT) for on-device robustness, and (v) temperature scaling for confidence calibration. Evaluations on the MycoAI-Lab and FieldMyco-Real datasets demonstrate that Deep SporeNet achieves 94.2% accuracy, 92.8% macro-F1, and a tail-class F1 of 82.7%, outperforming state-of-the-art mobile CNNs while running in < 50 ms on a standard smartphone processor. Its well-calibrated predictions (ECE = 0.039) and interpretable Grad-CAM visualizations confirm suitability for real-time fungal diagnostics, crop protection, and biodiversity monitoring. The model thus represents a scalable step toward AI-assisted mycology that is both data-efficient and deployable in resource-limited settings.

Deep Learning, Fungal Image Classification, Lightweight CNN, Few-Shot Learning, Mobile AI.

1. Krizhevsky, A., Sutskever, I., and Hinton, G. (2012). ImageNet classification with deep convolutional neural networks. Proceedings of the Neural Information Processing Systems (NIPS). [Google Scholar] [Crossref]

2. Sandler, M., et al. (2018). MobileNetV2: Inverted residuals and linear bottlenecks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 4510–4520. [Google Scholar] [Crossref]

3. Tan, M., and Le, Q. (2019). EfficientNet: Rethinking model scaling for convolutional neural networks. Proceedings of the International Conference on Machine Learning (ICML), 6105–6114. [Google Scholar] [Crossref]

4. Snell, J., Swersky, K., and Zemel, R. (2017). Prototypical networks for few-shot learning. Advances in Neural Information Processing Systems. [Google Scholar] [Crossref]

5. Finn, C., Abbeel, P., and Levine, S. (2017). Model-agnostic meta-learning for fast adaptation of deep networks. Proceedings of the International Conference on Machine Learning (ICML). [Google Scholar] [Crossref]

6. Sung, F., et al. (2018). Learning to compare: Relation network for few-shot learning. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1199–1208. [Google Scholar] [Crossref]

7. Vinyals, O., et al. (2016). Matching networks for one shot learning. Proceedings of the Neural Information Processing Systems (NIPS). [Google Scholar] [Crossref]

8. Cubuk, E. D., et al. (2020). RandAugment: Practical automated data augmentation with a reduced search space. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops). [Google Scholar] [Crossref]

9. Wang, D., Shelhamer, E., Liu, S., Olshausen, A., and Darrell, T. (2021). Tent: Fully test-time adaptation by entropy minimization. Proceedings of the International Conference on Learning Representations (ICLR). [Google Scholar] [Crossref]

10. Woo, S., Park, J., Lee, J.-Y., and Kweon, I. S. (2018). CBAM: Convolutional block attention module. Proceedings of the European Conference on Computer Vision (ECCV). [Google Scholar] [Crossref]

11. Howard, H., et al. (2021). Quantization and pruning for deep neural network compression. IEEE Access, 9, 15510–15525. [Google Scholar] [Crossref]

12. Zhang, X., Han, S., Mao, H., Hu, H., and Dally, W. (2018). ShuffleNet: An extremely efficient convolutional neural network for mobile devices. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). [Google Scholar] [Crossref]

13. Kumar, N., et al. (2023). Automated classification of microscopic fungal spores using deep learning. Computers in Biology and Medicine, 152, 106384. [Google Scholar] [Crossref]

14. Singh, R., and Sharma, P. (2024). Deep learning in plant pathology: Fungal disease recognition. Frontiers in Artificial Intelligence, 6, 112–120. [Google Scholar] [Crossref]

15. Fernandes, A., et al. (2024). Few-shot learning for rare pathogen identification in microscopy images. IEEE Transactions on Medical Imaging, 43(2), 304–318. [Google Scholar] [Crossref]

16. Lopez, P., and Joshi, M. (2023). Color normalization and illumination invariance in microscopic imaging. Pattern Recognition Letters, 171, 42–49. [Google Scholar] [Crossref]

17. Zhang, L., et al. (2024). Meta-learning for rare class detection in biological imaging. Bioinformatics Advances, 3(2), 210–222. [Google Scholar] [Crossref]

18. Jiang, G., and He, Y. (2023). Mobile AI for in-field plant disease diagnostics using quantized CNNs. Computers and Electronics in Agriculture, 205. [Google Scholar] [Crossref]

19. Dutta, S., et al. (2024). Explainable AI for microbiological image analysis: A review. IEEE Access, 12, 20403–20419. [Google Scholar] [Crossref]

20. Zhang, H., Wang, Y., and Zhao, X. (2025). Calibration of deep neural networks for reliable healthcare diagnostics. IEEE Journal of Biomedical and Health Informatics, 28(4), 2351–2363. [Google Scholar] [Crossref]