Rice Varieties Identification using Deep Learning

Rice Varieties Identification using Deep Learning

IEEE BASE PAPER TITLE:

Rice Variety Identification Based on Transfer Learning Architecture Using DENS-INCEP

IEEE BASE PAPER ABSTRACT:

Rice is a vital staple food for billions of people worldwide, especially in Asia, Africa, and Latin America, where it plays a key role in daily caloric intake and nutrition. With numerous varieties differing in size, shape, colour, texture, and nutritional content, accurate rice variety identification is critical for optimizing production and ensuring food quality.

Environmental factors such as soil type and climate further influence these variations, making precise identification essential for improving productivity and reducing waste. However, traditional manual methods of identification, relying on visual characteristics, are prone to human error, resulting in variety mixing, reduced quality, and higher costs.

This study addresses these challenges by employing the DENS-INCEP model, a transfer learning approach that integrates DenseNet-201 with the Inception module. DenseNet-201 serves as the backbone for feature extraction, while the Inception module enhances the model’s ability to capture multi-scale shape-related features, significantly improving classification accuracy.

The model achieved remarkable performance, with an average accuracy of 99.94% across multiple rice varieties. By implementing the DENS-INCEP model, this study contributes to Sustainable Development Goal (SDG) 2 by improving food security through enhanced rice production and supply chain stability.

Additionally, it supports SDG 9 by fostering innovation and advancing sustainable agricultural technologies. Furthermore, by reducing errors, waste, and inefficiencies in production and distribution, the model aligns with SDG 12, which emphasizes sustainable consumption and production. Overall, the DENS-INCEP model offers a robust and efficient solution to rice variety identification, addressing global food security challenges while promoting sustainability.

PROJECT OUTPUT VIDEO:

ALGORITHM / MODEL USED:

DenseNet-121, MobileNet.

OUR PROPOSED PROJECT ABSTRACT:

Rice is one of the most widely consumed staple foods across the globe, especially in Asian countries. Accurate identification of rice varieties is essential for improving quality control, seed selection, supply chain management, and ensuring food authenticity. Traditional methods of rice classification, often relying on manual observation, are time-consuming, error-prone, and inefficient for large-scale applications. To address these challenges, this project titled “Rice Varieties Identification using Deep Learning” proposes an automated, highly accurate classification system utilizing state-of-the-art deep learning techniques.

The system is developed using Python for backend logic, with a user-friendly web interface built using HTML, CSS, and JavaScript, and deployed using the Flask web framework. Two powerful convolutional neural network models, DenseNet-121 and MobileNet, are implemented and evaluated. The DenseNet-121 model achieved a training accuracy of 99.0% and a test accuracy of 99.3%, while the lightweight MobileNet model outperformed with a training accuracy of 99.4% and a test accuracy of 99.5%, making it an ideal candidate for real-time, resource-efficient applications.

The dataset used consists of 60,000 rice grain images, equally distributed across five classes: Arborio, Basmati, Ipsala, Jasmine, and Karacadag, with 12,000 images per class for training. The data preparation phase involved systematic steps including defining data directories, setting uniform image dimensions, image rescaling, structured loading, and preprocessing, ensuring high model generalization and accuracy.

The results demonstrate that the proposed deep learning-based system can reliably distinguish between closely related rice varieties with high precision. The project not only contributes to agricultural digitalization but also paves the way for scalable deployment in food quality assurance, automated grain sorting, and supply chain monitoring applications.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS: 

• System : Pentium i3 Processor.
• Hard Disk : 20 GB.
• Monitor : 15’’ LED.
• Input Devices : Keyboard, Mouse.
• Ram : 8 GB.

SOFTWARE REQUIREMENTS: 

• Operating System : Windows 10 / 11.
• Coding Language : Python 3.12.0.
• Web Framework : Flask.
• Frontend : HTML, CSS, JavaScript.

REFERENCE:

WAHYUDI AGUSTIONO, KURNIAWAN EKA PERMANA, CAROLINE CHAN, DESHINTA ARROVA DEWI, AND MOCH. MIFTACHUR RIFQI AL HUSAIN, “Rice Variety Identification Based on Transfer Learning Architecture Using DENS-INCEP”, IEEE Access, Volume: 13, 2025.

👉CLICK HERE TO BUY THIS PROJECT “Rice Varieties Identification using Deep Learning” SOURCE CODE👈

 

Frequently Asked Questions (FAQ’s) & Answers:

1. What is the purpose of this project?

The project aims to develop a deep learning-based system that can automatically identify different rice varieties—Arborio, Basmati, Ipsala, Jasmine, and Karacadag—based on images of rice grains. It helps in accurate classification for applications in agriculture, quality control, and food processing.

2. Which deep learning models are used in this project?

Two convolutional neural network (CNN) models are used: • DenseNet-121 • MobileNet Both models are implemented using TensorFlow and Keras, and they are trained to classify rice grain images with high accuracy.

3. What is the accuracy of the models used?

• DenseNet-121: 99.0% (Train), 99.3% (Test) • MobileNet: 99.4% (Train), 99.5% (Test) MobileNet achieved the highest accuracy while being lightweight and faster in inference.

4. What dataset was used for training the models?

A rice grain image dataset containing 60,000 images (12,000 images per class) across five rice varieties: Arborio, Basmati, Ipsala, Jasmine, and Karacadag. The dataset was preprocessed by resizing, rescaling, and structuring into train/test directories.

5. How can a user interact with the system?

Users can upload rice grain images through a web interface developed using HTML, CSS, JavaScript, and powered by a Flask backend. Once an image is uploaded and a model is selected, the system predicts the rice variety and displays the result instantly.

6. Is internet required to use this system?

No, the system can be run locally without the need for internet if hosted on a local server. However, for online deployment or cloud access, an internet connection is required.

7. Is the system suitable for industrial or commercial use?

While the system demonstrates high accuracy and efficiency, it is currently designed as an academic/final-year project. For industrial deployment, additional steps such as model robustness testing, real-world dataset expansion, and API integration would be needed.

8. How does this project differ from the IEEE base paper (DENS-INCEP)?

Unlike the complex and heavyweight DENS-INCEP model from the base paper, this project uses lighter models (DenseNet121 and MobileNet) that achieve similar or even better accuracy with less computational cost, making it more suitable for real-world, resource-constrained applications.

9. What are the main features of the system?

• Upload and classify rice images via web interface • Real-time rice variety prediction • High accuracy using deep learning • Support for two model choices (DenseNet121 & MobileNet) • Responsive and user-friendly UI