A Robust Deep Learning Segmentation and Identification Approach of Different Bangladeshi Plant Seeds Using CNN

Abstract

The purpose of this research is to identify the image of the seed. Different types of seeds are leveled into different classes through seed classification process. The demonstration was applied on more than 1000 seed images. It contains five processing modules such as Image acquisition, Pre-processing, Feature extraction, Image recognition and Show results. Seed varieties and qualities of seeds will be able to identify. The seeds species, namely Oryza sativa, Lagenaria siceraria, Cucurbita moschata, Zea mays, Benincasa hispida. Define the seeds age, shape, color, length, width, size, healthy or not, duration of seeds quality. The cost of analyzing the image will be minimal and does not require skilled labor, Seed studies--such as seed germination, can play an important role in seed purity. This research we have used CNN algorithm and the training accuracy was 87%-89% and validation accuracy was 90%-93%.