This project is a based on deep learning experiment, I focused on classifying images of three popular Pokemon: Pikachu, Meowth, and Bulbasaur, using Convolutional Neural Networks (CNNs). Built using Python and TensorFlow, it demonstrates how image recognition techniques can be applied to custom Image datasets.
The dataset contains labeled images of:
- Pikachu
- Meowth
- Bulbasaur
Each category includes a set of training and testing images.
- Python
- TensorFlow / Keras
- NumPy
- OpenCV
- Matplotlib
- Jupyter Notebook
- Image loading and preprocessing (resizing, normalization)
- CNN model building with Keras Sequential API
- Training with categorical cross-entropy
- Validation on test images
- Confusion matrix visualization
- Built a working CNN from scratch
- Applied classification to real-world image data
- Learned to fine-tune models for better accuracy