Monocots and Dicot Weeds Growth Phases using Deep Convolutional Neural Network

Authors

  • Anand Muni Mishra, Vinay Gautam

Abstract

The agriculture plays vital role in societies and requires research, planning and execution. In this research paper we have classified 20 types of weed species growth that grow in the Ravi season .It is impotent to identify and growth estimation of weed using deep learning technology in the field of convolution neural networks. This looks at situ images involving 20 weed species grown within a Time, 10500 leaves of these drawings were used for the trained of the weed statistics is taken from the rabi crop and its weeds are divided into 20 increase instruction. Image was an ordinary overall performance assessment of this proposed convolutional neural network using inceptionv3 model on 2000 pictures tested, which in addition are several in cropping, soil types, photograph judgments, and lighting fixtures situations. The common ordinary performance of this method met the maximum accuracy of 90.79% and the minimum accuracy of 25.68% for detecting polygon images. Minimum accuracy < 8 count of leaf for amaranthaceous family weeds. Further, it carried out a mean 90% accuracy charge whilst estimating the Leaf deviation accepting wide variety of leaves and 84% accuracy > 8 leaf count leaf. The dataset consists of 1500 images identifying 6 weed households with 20 weed species, in overall 10500 annotations have been made. Three RGB virtual cameras have been used for photo capturing: Intel Real Sense LiDAR digital camera L515, Canon digital SLR DIGICAM EOS 850 D 18-55IS STM, and SONY w800. Pictures had been grown on food crops and weeds in a managed environment and taken at discipline conditions at specific increase degrees the ones consequences propose that The CNN and its associated models can process a lot of weed pictures simultaneously.

Keywords- Convolutional Neural Network; inception v3; Weed Growth Stage; weed leaf Counting;computer vision;

Published

2020-12-11

Issue

Section

Articles