Potato Disease Detection using Deep Learning

Overview

Farmers who grow potatoes face significant economic losses due to diseases affecting potato plants. Two of the most common diseases are:

• Early Blight (caused by a fungus)
• Late Blight (caused by a microorganism)

Timely detection of these diseases can help farmers take appropriate actions and prevent crop losses. AtliQ Agriculture, an AI company, has taken the initiative to develop a mobile application that allows farmers to detect these diseases by simply capturing an image of the potato plant. The app utilizes Deep Learning and Convolutional Neural Networks (CNN) to classify the plant as:

• Healthy
• Early Blight Infected
• Late Blight Infected

Features

1. Identify whether the potato plant is healthy or diseased
2. Detect Early Blight and Late Blight
3. Uses Convolutional Neural Networks (CNN) for classification
4. Built for mobile integration
   5.Helps in reducing economic losses for farmers

Dataset

The dataset used for training the model consists of images of healthy, early blight, and late blight potato plants. It has been collected from various agricultural sources and processed for training. The dataset is taken from Kaggle through this link.

Model Architecture

The model is built using TensorFlow/Keras and follows a CNN-based approach:

1. Preprocessing: Image resizing and normalization
2. Data Augmentation: Random flipping, rotation to improve generalization
3. CNN Layers:
   • Convolutional layers with ReLU activation
   • MaxPooling layers for feature extraction
   • Fully connected (Dense) layers for classification
4. Softmax Output Layer: Outputs probabilities for the three classes

Training & Performance

• Optimizer: Adam
• Loss Function: Sparse Categorical Crossentropy
• Evaluation Metric: Accuracy
• Training done using Google Colab