Collaborative Development: Worked in a team of three—defining layers, pooling, and dropout to combat overfitting.

Data Pipeline: Processed 60,000 labeled images into 5 classes with real-time augmentation for robust training.

Model Metrics: Achieved accuracy 68.79%, precision 70.13%, recall 68.79% and F1 69.17%.

Tech Stack: Built in Python with TensorFlow, tracked experiments in VS Code, and visualized results via matplotlib.