Detecting Bangladeshi flowers and its classifications from image using CNN

Abstract

Image processing has been one of the most practiced features of digitalization and it has enormous role in different sectors of our life. Convolutional Neural Network (CNN) has made image processing and other mechanism related to images much more significant and glorious. We captured images of flowers in different angles with different shades and nature mode so that while analyzing it can call on enormous data even if candid image is taken in testing purpose. We went through MobileNet, InceptionV3, and VGG16 models to find the best suited one and that will be automatically called by the system and with that algorithm the rest analysis will occur. We found 95.39% accuracy for MobileNet. The rest InceptionV3 and VGG16 have 94.68% and 92.24% accuracy respectively. This will be very much beneficial to know about our native nation and its resources and such dataset is rarely found on web when we insist about Bangladesh.