Data Collection & Processing:

​

• Data Source: Images of healthy and wilted houseplants were sourced from the Kaggle dataset Healthy and Wilted Houseplant Images.

• Data Preparation: The images were split into training and test sets with an 80/20 split to ensure proper model validation. All images were resized to 256x256 pixels for consistency.

• Handling Imbalanced Data: I reviewed the class distribution to check for imbalances between the healthy and wilted plant categories, finding that the data was fairly balanced, with no need for additional handling techniques.

​

Data Analysis:

​

• Conducted exploratory analysis on the dataset to inspect the distribution of images across the "healthy" and "wilted" categories.

• Visualized the class distribution using bar charts to ensure balanced data.

• Randomly sampled images from both classes to get a qualitative feel for the dataset and to understand the variety in plant images.

​

Model Training:

​

• Utilized TensorFlow to load and preprocess the image dataset, splitting it into training and validation sets.

• A CNN-based machine learning model was built to classify the houseplants into the two categories.

• Implemented techniques such as image augmentation and batch processing to improve model performance and avoid overfitting.

​

Visualization & Reporting:

​

• Visualized class distribution using Python’s matplotlib.

• Displayed samples of images to illustrate the appearance of healthy versus wilted plants.

• Developed a comprehensive report documenting the data preprocessing steps, model architecture, and performance metrics, including accuracy and validation results.