Nutrients Deficiency Detection in Crops using Machine Learning
This project aims to detect the nutrient deficiency in crops using the techniques of machine learning. Deficiency of nutrients in plants result in a reduced crop yield or low plant quality. Even if symptoms are encountered, time investment and analytical costs are still required.
To deal with this, our system is expected to provide an easy solution to perform a timely diagnosis of nutrient deficiency. We are performing the deficiency analysis on the images of rice leaves belonging to four classes i.e. Nitrogen Deficient, Potassium Deficient, Phosphorus Deficient and Healthy leaf. The dataset is very limited so we need techniques to cope up with data limited situations.
For this purpose, we have used transfer learning where 25 layers of ResNet50, built for ImageNet classification, are used as it is and the remaining layers are re-trained for our dataset. One shot learning is another useful technique where we have trained the model on large number of similar and non-similar pairs of the images after which the predictions are done by calculating the Euclidean distance between them.
We have evaluated the performance of the model using precision, accuracy, recall and F1 score and obtained satisfactory results on unknown data.
The final step is to develop a system for users to perform the detection. For this purpose, we are developing a mobile application using Android Studio. It will take Leaf image from the user which will then be sent to the cloud where the built Deep Learning model will be deployed on it and the deficiency results will be sent back and displayed to the user.
Keywords: Nutrient Deficiency, Machine Learning, Mobile Application
Tools: Tensorflow, Keras, OpenCV, NumPy, Android Studio
Department: Department of Business Studies
Project Team Members
| Name |
Email |
|
Haider Chaudhry
|
haider2018@namal.edu.pk |
|
Shanze Nawaz Khan
|
shanze2018@namal.edu.pk |
Project Poster
