Comparative Analysis of Computational Paradigms in Food recognition and Classification

This paper presents recent improvements in the field of food image classification and demonstrates the efficacy of deep learning models in performing classification tasks. The current paper studies the following models: (i) a scalable model for food classification built on PyTorch and PReNet (ii) an Indonesian food recognition model optimized for mobile devices with EfficientNet-Lite and (iii) an Indonesian cuisine classification model based on pre-trained CNNs (Inception-V3).   The PReNet model is one of the most effective models when it comes to large-scale classification with its Top-1 accuracy of 84.03% and Top-5 accuracy of 98.21%.The mobile-friendly EfficientNet-Lite model performs at the level of 97.78% accuracy, which makes it useful in real-time applications even though its training dataset is smaller. Inception-V3 is focused on efficient classification of Indian cuisine while integrating calorie and nutrition detection.   Challenges associated with the interpretability of the models, diversity of datasets, and real-time performances need to be addressed in future research. There is a need for optimization of processing speed, expansion of training datasets and development of explainable AI models.   In conclusion, AI food recognition models have a significant potential for usage in healthcare, diet tracking, IOT applications and other domains.

[1]           Min, W., Jiang, S., Liu, L., Rui, Y., Jain, R., & Qi, G. (2023). Large scale visual food recognition. *IEEE Transactions on Pattern Analysis and Machine Intelligence, 45*(8), 9932–9949. https://doi.org/10.1109/TPAMI.2023.3237871. [2]             Zhang, H., et al. (2021). Food recognition with fusion of deep convolutional features and hand-crafted features. IEEE Access, 9, 10193–10202   [3]             Liang, Y., et al. (2021). Food recognition with multi-viewpoint images. IEEE Transactions on Multimedia, 23, 1286–1298 .   [4]           Min, W., Jiang, S., Liu, L., Rui, Y., & Jain, R. (2020). A survey on food computing. ACM Computing Surveys, 52(5), 1–36.   [5]             Min, W., Wang, Z., Liu, Y., Luo, M., Kang, L., Wei, X., Wei, X., & Jiang, S. (2023). Large scale visual food recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(8), 9932–9949. https://doi.org/10.1109/TPAMI.2023.3237871.   [6]               Wei, X., Jiang, Y., & Liu, S. (2021). Food image recognition with multi-modal learning and attention mechanisms. IEEE Transactions on Multimedia, 23, 3305–3317. https://doi.org/10.1109/TMM.2021.3057896.   [7]             Subhi, M., Hamid, S., & Mohammed, M. (2019). Vision-based approaches for automatic food recognition and dietary assessment: A survey. IEEE Access, 7, 180311–180326. https://doi.org/10.1109/ACCESS.2019.2904519.   [8]             Min, W., Jiang, S., Liu, L., Rui, Y., & Jain, R. (2020). ISIA Food-500: A dataset for large-scale food recognition via stacked global-local attention network. ACM Multimedia, 1–9. https://doi.org/10.1145/3394171.3414031.   [9]             Min, W., Liu, L., Luo, Z., Jiang, S., & Wei, X. (2019). Multi-scale multi-view deep feature aggregation for food recognition. IEEE Transactions on Image Processing. https://doi.org/10.1109/TIP.2019.2929447. [10]             Zheng, J., Wang, Z. J., & Zhu, C. (2017). Food image recognition via superpixel based low-level and mid-level distance coding for smart home applications. Sustainability, 9(5), 856. https://doi.org/10.3390/su9050856.