Overview / Usage

plant seedling classification wherein we have to distinguish among 12 species of crops is the main aim of the project. This project intends to solve the problem using transfer learning, which aims to build an accurate model with less computation cost. a problem such as a crop identification plays an important role in disease identification and nutrients being supplied to the plant. Also visually impaired farmers can use this model to distinguish between various crops present in the field.

Identifying the crop also becomes necessary in case of planting them back in case their shortage is recorded. This model will, therefore, give an accurate classification of the crop in the field and help the farmer identify the pant he needs to plant back.

Methodology / Approach

Transfer learning is used to solve the above problem. In transfer learning, we alter the last few layers of pre-trained networks and add an additional layer for custom detection as per our dataset. here I have used RESNET50 as a backbone and have added an extra layer for the correct prediction of the crop.

Neural networks are trained through backpropagation wherein data set is fed to the network and it learns the image such that the cost function is brought to a minimum, till that time the weights are updated in the network. and the last layer of softmax function yields the probability value to which class the image may belong to.

Technologies Used

Keras

open cv

NumPy

matplot lib

flask

rest API

curl

TensorFlow

anaconda

python