From Field to Data: A Machine Learning Approach to Classifying Celery Varieties

This study investigates a comprehensive scheme that incorporates Convolutional Neural Networks (CNN) and K-Nearest Neighbors (KNN) for the classification of celery types, that is, ‘Pusa Jyoti’, ‘Utah’, and ‘Pusa Lehar’. Leveraging on the joint of CNN for feature extraction and KNN for identification of crop varieties harnesses the advantage of both algorithms intending to increase the precision of crop identification which is an essential component of precision agriculture that further improves crop management, yield quality, and sustainability. The paper puts forward a complete methodology for data preparation and cleaning, model configuration and tuning, and lastly model evaluation. The results of the experiments with a fully integrated CNN - KNN model are appraised using metrics like accuracy, precision, recall, and F1 score, which show that the performance is substantially better than the traditional methods. The research shows clearly the model's power to categorize celery species accurately, however, a comparative analysis with the existing state-of-the-art model has been done and the results show the superiority of the model. The investigation of this study goes beyond the mere academic contribution, it can serve to provide knowledge that applies to agricultural technology; hence farmers, agronomists, and the whole of the agricultural supply chain can benefit. While the research faced challenges that included developing comprehensive datasets that are easily available and numerous, this study provided a foundation for future studies and possible implementation of other machine learning algorithms that aim at improving the accuracy and applicability of the model to other crop types.