Germination Prediction System for Rice seed using CNN Pre-trained models

• Published in: 2022 International Conference on Advances in Computing, Communication and Applied Informatics (ACCAI) pp. 1 - 9
• Main Authors: Durai, S., Mahesh, C., Sujithra, T., Shyamalakumari, C.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.01.2022
• Subjects: Agriculture; Biological system modeling; Computational modeling; Computer vision; Crops; Germination; Paddy; prediction; Predictive models; Productivity; rice seed; Seed Quality; Training
• DOI: 10.1109/ACCAI53970.2022.9752611

In south India, rice is the most important food source, with 90 percent of people ingesting rice products on a daily basis. In agriculture nowadays farmers are facing lot of problems related to productivity. Even government is Rice is a major food source in south India. Majority of the people consuming rice products in day to day life, it creates a high demand. They are not able to attain the yield as expected because of climatic changes, shortage of water, and unsuitable crop selection irrespective of the soil nutrients etc., Agricultural field is well suited for the particular crop, and choice of seed will affect the crop productivity. It results huge loss in agriculture. Hence, seed selection plays a vital role in achieving good productivity. Nowadays, the people who want to do the agriculture they do not have the enough knowledge in seed/ crop selection. It results in less productivity. In order to address the issue, we proposed germination prediction system for rice seed using pre trained CNN models. The main objective of the proposed system is to make anyone to do the agriculture with the help of computer vision system. It gives assistance in seed selection. It increases productivity. It also creates a platform to the people who wants to do the agriculture without any agricultural background. We have given simple and economically feasible solution for predicting the germination of four different rice seeds namely AtchayaPonni, AndhraPonni, KO50 and IR20 which major varieties cultivated by tamilnadu farmers. For experimental analysis, We applied CNN with pre-trained models such as Alexnet, Resnet, inception _ v3.0 _prediction and got 89% 83%, and 98.44 % of accuracy respectively. Real time images of the rice seed is directly taken from agricultural land located in trichy for training and testing the models.