CNN-KNN Hybrid Model for Accurate Celery Variety Classification in India: A Three-Class Analysis

The proposed study shows a strong system to sort celery types in India using the Convolutional Neural Network (CNN) and K-Nearest Neighbors (KNN) model. The model used 8900 images of Pusa Jyoti, Utah, and Pusa Lehar celery types which helped it to get a very good total accuracy of 97.38% when recognizing different visual aspects. The data was carefully cleaned up, including making it all the same and picking out important features. This improved how well the model could see little differences between different types of plants. The evaluation measures - accuracy, finding rate, and F1 score showed the model did a good job with few mistakes across all types. Figure 3 shows a comparison with other methods. The CNN-KNN model for sorting celery types ranks as best practice, beating existing top standards when it comes to accuracy. However, problems about exceptions and math difficulties were mentioned. This shows chances for more improvements. This study shows that the model can be very helpful in farming. It helps to correctly identify plant types for best crop care practices. The model's ability to work in different situations is good for using it outdoors. While learning more about getting bigger and handling unusual values is something for the future, this study shows an important step in crop classification using computer teaching methods. The made-up CNN-KNN model can change how farming is done. It makes it easier for people involved to make smart choices about what they do on farms, replacing old methods with new ones that just work better.