Crop Recommendation using Machine Learning and Plant Disease Identification using CNN and Transfer-Learning Approach

Since there have been climate changes that have resulted in an increasing amount of unexpected rainfalls, par below temperatures, and heatwaves in the region, resulting in a significant loss of ecosystem. Machine learning has helped develop various utility tools to tackle world problems. This problem of agriculture can be solved by using various ML algorithms. This paper aims at two things - a)A crop recommendation system and b) a Plant disease identification system embedded into a single website. The datasets were publicly available over the internet. Once the features for task one are extracted, the dataset is trained on five different algorithms - logistic regression, decision tree, support vector machine(SVM), multi-layer perceptron and random forest. For the second task, three CNN architectures, VGG16, ResNet50 and EfficientNetV2, are trained, and a comparative study is done between them. For task one, random forest achieved an accuracy of 99.31%, and for the second task, EfficientNetV2 achieved an accuracy of 96.06%