Institution: Daffodil International University (DIU)

Role: Individual Researcher

Supervisor: Mr. Md. Sazzadur Ahamed (Assistant Professor)

Co-Supervisor: Mr. Md. Assaduzzaman (Lecturer, Senior Scale)

Developed an AI-based system to automatically grade dragon fruits as fresh or defective using machine learning and deep learning techniques.

• Collected and preprocessed a custom dataset of 5,011 images, applying data augmentation techniques
• such as flipping, zooming, and rotation.
• Designed and implemented multiple deep learning models—including a custom CNN, ResNet-50 (transfer
• learning), and EfficientNet-B0—using Python and TensorFlow/Keras.
• Achieved 99.5% validation accuracy and an AUC-ROC above 0.95 with EfficientNet-B0, demonstrating a low-cost,
• scalable AI-driven grading solution for agricultural quality assessment.