Pest Classification with CNN and Grad-CAM

Machine Learning · Computer Vision · Deep Learning

This project focuses on classifying pests from image data using deep learning. By applying CNN-based modeling and Grad-CAM visualization, the system not only predicts pest categories but also improves interpretability by showing which regions of the image influenced the model’s decision.

Key Highlights

• Built an image classification model to identify multiple pest categories.
• Used CNN-based deep learning techniques for feature extraction and prediction.
• Applied transfer learning to improve performance on image data.
• Implemented Grad-CAM visualization to interpret model predictions.
• Improved model understanding by analyzing attention regions and misclassifications.

What I Learned

• How deep learning models can be applied to real-world image classification problems.
• The importance of model interpretability in machine learning workflows.
• How Grad-CAM helps explain CNN predictions visually.
• How to evaluate and improve model behavior using visual feedback.