Abstract:
This research titled as “A Comprehensive Study on Multi-Class Classification and Damage
Identification in Bangladeshi Fruits using Deep Neural Networks” presents a
comprehensive exploration of fruit classification focusing on Bangladeshi local bananas,
employing deep learning techniques with a specific emphasis on the DenseNet201 model.
The study introduces a meticulously curated dataset, addressing the scarcity of banana
image data in the agricultural domain. Leveraging data augmentation techniques, the
dataset is expanded and utilized for training and evaluating the proposed transfer learning
model. The experimental setup involves robust hardware configuration and software
requirements, ensuring meticulous evaluation. The DenseNet201 model is proposed,
showcasing exceptional accuracy of 98.76%. Performance metrics, confusion matrices,
and training/validation curves provide a detailed analysis of the model's effectiveness. The
research discusses the impact on society, environment, ethical aspects, and outlines a
sustainability plan. The study concludes with implications for further research, highlighting
the dynamic nature of deep learning applications in agricultural technology.
