Rice Variety Identification Based on Transfer Learning Architecture

Project Code :TCMAPY1955

Objective

This project develops a rice variety identification system using transfer learning to classify rice grain images accurately and efficiently. Four advanced architectures—EfficientNetV2-S, EfficientNetV2-M, CNN, and Mobilenet v2—are fine-tuned to extract meaningful visual features such as texture, shape, and structural patterns. The models are trained and evaluated on a structured rice image dataset to compare their performance and determine the most effective classifier. A Flask-based web interface is created, enabling users to register, log in, upload images, and receive classification results in a simple and accessible manner. The system focuses on controlled, dataset-based classification and supports research-oriented analysis of transfer learning models. Overall, the project demonstrates how modern deep-learning architectures can provide accurate and efficient grain recognition, reducing reliance on manual identification and expert knowledge.

Abstract

This project presents a rice variety identification system built using transfer learning methods to enhance accuracy and reduce training complexity. The system employs four advanced architectures—EfficientNetV2-S, EfficientNetV2-M, MobileNetv2, and CNN—to classify rice grain images from a structured dataset. Each model is fine-tuned to extract detailed visual patterns related to texture, shape, and structural features. A Flask-based interface connects the trained models with a simple user platform containing modules for registration, login, classification, and logout. The aim is to create a lightweight and efficient classification system that supports research-oriented analysis and controlled image identification tasks. Experimental results indicate strong performance across the chosen architectures, offering reliable identification capabilities that can be extended to various controlled environments. This work demonstrates the usefulness of transfer learning for efficient feature extraction and improved classification accuracy in image-based grain recognition tasks.

Keywords: Transfer Learning, EfficientNetV2-S, EfficientNetV2-M, MobileNetv2, CNN, Image Classification, Rice Identification, Deep Learning, Flask Framework, Feature Extraction

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.