Research Projects

Rice is globally a key economic driver for millions of farmers. It is regularly eaten making it one of the important staple foods for over 3.5 billion people. Rice Grain Quality Analysis is essential to global food supplies. However, traditional methods of rice quality assessment, which rely on manual inspection are labor-intensive, costly, and time consuming. Therefore, we believe that to make the process faster and cost effective, deep computer vision models can be used to automated quality assessment of rice grain. Our approach is based on image analysis to facilitate classification and counting tasks of rice grain. We plan to utilize the visual rice grain characteristics such as size, shape, color, and texture to build a strong feature extractor that will attain high accuracy for the downstream tasks. We plan to use Self-Supervised Representation Learning (SSRL) to solve the rice grain analysis problem. Recent research in computer vision highlights the capability of convolutional networks to outperform traditional methods for object detection, segmentation and counting tasks. The traditional supervised approaches require a huge amount of labeled data to support fine-grained classification. In contrast, SSRL approaches allow one to build a good feature extractor without requirement of labels. The developed feature extractor can support various downstream tasks such as classification, segmentation, counting, and is also robust to variations in imaging conditions. The project also aligns with various government initiatives, such as Digital India, AI for All, and the Doubling Farmers Income. We believe the proposed project addresses the important concerns in export quality and provides essential tools for agricultural technology development and economic growth.