Project Showcase
Highlight of my featured works
🖼️✨ Autoencoder for Image Denoising
A convolutional autoencoder model trained to remove noise from images. Synthetic noisy images were generated to simulate real-world denoising challenges.
🎯 What I Learned
- Built convolutional autoencoder for image denoising
- Generated synthetic noisy datasets for training
- Applied dropout and tuning for sharper reconstructions
🛠 Tech Stack
- Python, NumPy, Matplotlib
- OpenCV, PIL
- Scikit-learn
- TensorFlow / Keras, Keras Tuner
📌 Evaluation / Next Step
The tuned autoencoder reduced reconstruction error (MSE) compared to the baseline and produced sharper denoised images. However, since the training used synthetic noise, generalization to real-world noisy images may be limited. Future improvements include using real noisy datasets, exploring perceptual loss functions, or adopting U-Net-based architectures.