This project implements a Convolutional Autoencoder using TensorFlow/Keras to perform image compression and reconstruction on the MNIST handwritten digits dataset.
This project implements a Convolutional Autoencoder using TensorFlow/Keras to perform image compression and reconstruction on the MNIST handwritten digits dataset.
- Implemented Encoder–Decoder CNN architecture with Conv2D, MaxPooling2D, and UpSampling2D layers.
- Trained on 50,000+ MNIST images for unsupervised feature learning.
- Visualized reconstructed images vs original images for performance evaluation.
- Demonstrates understanding of deep learning, data preprocessing, and model evaluation.
- Python, NumPy, Matplotlib
- TensorFlow / Keras
- Scikit-learn
python autoencoder_mnist.py