dc.description.abstract |
Automatic fruits classification is becoming more popular day by day. Many of us have
very few idea about local fruits so we can attest to them even if we don’t know about the
local fruits. In our thesis, we will describes about a system that can automatically
recognize local fruits using computer vision approach. It is very common process to
identifying fruit however automatic fruit classification is not easy task depending on their
object’s positions, shape, colors, etc. In our project, we have collected the samples from
different local area and then we applied different deep learning models like Resnet-50,
VGG-19, Inception-v3, MobileNet, etc. that used Convolutional Neural Network (CNN)
techniques to detect local fruits and classify in different classes. Among them MobileNet,
VGG-19 given us higher accuracy of 99% and 98%. We also proposed a best model
based on our training accuracy.
We have collected eight different types of local fruits to done the project. We have total
3240 samples among them for training purpose we used 60 percentage of image data, 20
percentage of image for validation and 20 percent of total image used for testing purpose.
To get better result, we removed image background and then augmented them in various
way. After training and testing we got satisfied result. As the result of this research model
local fruits detection are classified, which can help in our daily life to identify local fruits |
en_US |