Recognition of Cotton Crop Diseases and Treatments using Deep Learning

Project Title

Recognition of Cotton Crop Diseases and Treatments using Deep Learning

Project Area of Specialization Artificial IntelligenceProject Summary

Many studies show that quality of agriculture products depends upon many factors which an either causes increase or decrease. One of the most important factors to maintain such quality of plant is by Consequently, minimizing plant diseases which allows to keep improving quality and quantity of the production. The project is based on the Artificial Intelligence which will keep learning with its every use. In times like these where agricultural production is decreasing due to many reasons like climate change new type of pest etc., we need new innovative ideas to increase its yield. Cotton Care is also one of the steps to integrate artificial intelligence into the agriculture. 

As we know that human eye is not proficient enough to detect some diseases correctly because even some minute variation like pattern or color can be a whole different disease present on the cotton crop leaf. Our software can exactly differentiate the difference of color or pattern present on these leaves and depending upon that difference it could further compare it with image features related to the colors and patterns already stored in database.

The model is trained to achieve intelligent farming, including early detection of disease in the leaves, fruit and stem etc. So, our model is giving farmers an easy and efficient method to detect the cotton diseases and pests affecting it, and it will also recommend them what type of pesticides to use to overcome the disease. This work is based on Deep Learning model in which, the captured images are processed for enrichment first, then texture and color. Feature extraction techniques are used to extract features for the detection and recognize of the diseases and after the disease is recognized it will further tell what kind of pest is affecting it.

Project Objectives

• To select appropriate cotton crop real time data for training and processing.
• To detect cotton crop diseases using DL
• To identify pests, harming the cotton crop health using DL
• To recommend the treatment for detected crop disease and pests.

Project Implementation Method

After we are done making the model, we will train our model as shown on the right side of the flowchart we have collected all the data that is required and stored it in the storage then from the storage we will process the data in which the data will resized and reshaped according to requirement our model.

After the data is processed it will pass through feature extraction process in which its edges, texture, color, shape, and background features will be extracted after feature extraction we will split the data in 80%, 20% partition, 80% for training and 20% for testing. Transfer learning is used to retrain the inception model of Google to train CNN for cotton crops, we will freeze the variables of other layers of the inception model so it’s already trained layers do not get messed up and lose their accuracy, as we are using transfer learning to increase our model’s accuracy, we will train only the last layer of our model with our training and testing data for a certain number of epochs using adam optimizer. After we are done with training, we get our trained model ready for prediction.

In the left side of our flowchart our prediction model is shown. Now we will provide it with appropriate image which will go through image processing to reshape and resize the image then in second step its features will be extracted after that the image will be given to the model for label prediction after we get the predicted label the rest of the information will be given with the label.

images/Recognition of Cotton Crop Diseases and Treatments using Deep Learning _1639949696.png

images/Recognition of Cotton Crop Diseases and Treatments using Deep Learning _1639949698.png

Benefits of the Project

• Commercial:
  • Will help companies to sell their product online so it will not go through third party distributers which will help in keeping product at most affordable prices.
• Farming:
  • Project will help in increasing the production of the healthy crops.

• Cost saving
  • The main reason for decrease in cotton crop is its high production cost. So this project will help in giving farmers more cost effective way of online pesticide purchases from companies.

• Time - saving
  • Traditional diagnosing method is a lot time consuming. In traditional method, Diagnoses is done by field officers, who come from different cities and have many fields in line to visit. Till field officers visit the field, many crops would be effected by disease till then. As are some diseases that can destroy the crop in one day even before the field officer come to visit for diagnoses.  

• Higher Accuracy:
  • Accuracy is the not certain in manual diagnosis as manual diagnosis is done by field officers who can make mistakes due to many factors. But in this project accuracy issue taken into context and a machine learning-based model is developed to overcome all previously described accuracy issues and will certainly provide better results than human can.

• Cure Recommendation:
  • Along with determining diseases of crops, our model will give the best suited treatment/pesticide required for our diseased plant. 

• More production:
  • All the upper factors will ultimately effect the production positively, which will be beneficial for farmers as well as for a country’s GDP.

• Research:
  • Will help other student who are working on cotton crops by giving the valuable information and data.

Technical Details of Final Deliverable

Our final product will be a software-based application which will process captured images of cotton crops. The model will process the image for enrichment first, then texture and color Feature extraction techniques are used to extract features such as boundary, shape, color and texture for the disease spots to recognize the diseases, along with disease detection, it will also recognize the pest that causes disease. After recognizing diseases and pests for cotton crop, our model will finally provide a certain pesticide formula that will help to get rid of the disease on the cotton crop.

For all the detail discussed above, we will be using deep learning-based algorithms.

We will take real time image of cotton crop and will provide it to our modelled software which will first process the image and make it suitable for the further processing. Then the image will go to our trained model which will then process the given data and predict disease, pest and cure formula for the crop.

Final Deliverable of the Project Software SystemCore Industry AgricultureOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Decent Work and Economic Growth, Responsible Consumption and Production, Climate Action, Life on LandRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	69750
Camera	Equipment	1	40000	40000
Cloud gpu	Equipment	15	1250	18750
Magnifying glass	Equipment	2	1500	3000
Traveling	Miscellaneous	1	8000	8000