Improved Celery Variety Classification in India Using CNN-KNN Hybrid Model

Classification of celery varieties plays a primary role as an essential factor of agriculture that works towards increasing productivity and sustainability. Age-old methods, though at times laborious and error-prone, are sometimes, the most efficient and reliable. This study reveals a fresh perspective on celery variety classification in India, employing a Human Hybrid Convolutional Neural Networks (CNN) - K-Nearest Neighbors (KNN) architecture to blend the strengths of deep learning and instance-based learning. This integration will prominently enhance the accuracy of results. A uniform dataset of celery images was created including a wide diversity of growth stages and conditions; this dissimilar dataset has been applied to this CNN-KNN hybrid model. Our model was trained, developed, and tested rigorously, therefore it showed a remarkable improvement (94%) in classification accuracy in this respect to existing methods. The discovery validates the work of the hybrid model in correctly classifying the celery varieties giving huge progress to the machine learning (ML) use in farming. The study demonstrates that the cassette network and k-nearest neighbor approach are promising for crop variety classification, which motivates future research into precision agriculture methods. The study not only provides a hint for better management of similar technologies but also indicates the usefulness of the hybrid ML models in pressing issues in the agricultural sector.