Bittergourd Classification

Maturity Prediction Of Bitter Gourd Using bitter-net

This research addresses the need for efficient and accurate classification of bitter gourd images into three distinct classes Immature, Partially matured, and Matured. The current lack of automated solutions blocks agricultural automation and effective crop management. To overcome this challenge, a comparative approach employing several Convolutional Neural Network (CNN) models is proposed. The selected models, including VGG16, MobileNet, InceptionNet, AlexNet, ResNet and DenseNet, offer diverse architectures, allowing for an in-depth evaluation of their effectiveness in bitter gourd classification. VGG16 is recognized for its simplicity and complexity, while MobileNet's lightweight architecture makes it exceptionally efficient. DenseNet promotes dense connectivity for increased feature reuse, AlexNet is known for its groundbreaking design that pushed forward deep learning, ResNet is famous for its ability to train very deep networks effectively, and InceptionNet offers inception modules for improved feature extraction. The evaluation criteria consist of three factors: recall, precision, and F1-score. These factors together provide information about how well the models can classify images of bitter gourds at different stages of development. Furthermore, the model's predictive accuracy is quantitatively evaluated by the Root Mean Square Error (RMSE), which measures the average divergence between the expected and actual maturity phases. Additionally, the classification results are visualized using the confusion matrix, which allows for a thorough analysis of the model's abilities to distinguish between the stages of bitter gourd. The solution involves training these models on a selected dataset of bitter gourd images, with labels corresponding to different maturity stages. The anticipated outcome of this research is to identify the most accurate and efficient deep-learning model for bitter gourd classification. And fine-tuning that model to a new proposed model called Bitter-Net. By doing so, I contribute to the advancement of agricultural automation and crop management.