"CROPCARE AI": AN END-TO-END DEEP LEARNING SYSTEM FOR REAL-TIME POTATO DISEASE DETECTION AND CLASSIFICATION

CropCare AI has been shown through this paper to be an example of a functional platform for the detection of diseases affecting potato leaves by leveraging deep learning models as well as the entire infrastructure needed for the deployment thereof[15]. An image classification problem based on PlantVillage data was solved using CNNs, and this was deployed in an ensemble system comprising a FastAPI backend inference server, React Native mobile app, and serverless cloud functions on GCP. This project shows how deep learning can move from the lab to a real-world application. The deployment has focused on the specific problem of early detection of diseases affecting potatoes, a crop for which early intervention is paramount in order to limit losses. The hybrid deployment method used has allowed the tool to be deployable even in areas with limited internet access, where the solution would be applied.   A few research directions for future study are mentioned below. Expanding the classification list of diseases and plants will improve the efficacy of the system greatly from an agronomic point of view. The inclusion of geospatial information together with image information can help generate regionalized advice based on disease prevalence trends in that specific region. Regular retraining of the models with field images collected at different times can lead to improvements in their robustness with respect to varying imaging conditions. Also, implementing multilingual capability for generating advice will facilitate the use of the system by many other farmers.

[1] S. P. Mohanty, D. P. Hughes, and M. Salathé, "Using Deep Learning for Image-Based Plant Disease Detection," Frontiers in Plant Science, vol. 7, p. 1419, Sep. 2016. [2] D. P. Hughes and M. Salathé, "An Open Access Repository of Images for Identification of Plant Diseases," arXiv preprint arXiv:1511.08060, 2015. [3] A. Kamilaris and F. X. Prenafeta-Boldú, "Deep Learning in Agriculture: A Survey," Computers and Electronics in Agriculture, vol. 147, pp. 70–90, Apr. 2018. [4] J. G. A. Barbedo, "A Review on the Main Challenges in Automatic Plant Disease Identification Based on Visible Range Images," Biosystems Engineering, vol. 144, pp. 52–60, Apr. 2016. [5] D. P. Kingma and J. Ba, "Adam: A Method for Stochastic Optimization," in Proc. Int. Conf. Learning Representations (ICLR), San Diego, CA, USA, 2015. [6] TensorFlow Developers, "TensorFlow: An Open-Source Machine Learning Framework," TensorFlow.org, 2023. [Online]. Available: https://www.tensorflow.org [7] S. Ramirez, "FastAPI: Modern, Fast Web Framework for Building APIs with Python 3.6+," GitHub, 2023. [Online]. Available: https://fastapi.tiangolo.com [8] Google Cloud, "How to Serve Deep Learning Models Using TensorFlow 2.0 with Cloud Functions," Google AI Blog, 2021. [Online]. Available: https://cloud.google.com/blog/products/ai-machine-learning [9] Meta Open Source, "React Native — Build Native Apps Using React," reactnative.dev, 2023. [Online]. Available: https://reactnative.dev [10] TensorFlow Team, "TensorFlow Lite: On-Device Machine Learning," TensorFlow.org, 2023. [Online]. Available: https://www.tensorflow.org/lite