This is a lightweight jupiter notebook of my Unsupervised Image Clustering project based on PCA, NMF, Autoencoder and DEC.
This repository contains a lightweight version of my unsupervised image clustering project.
It provides a clear, high-level view of the methodology, the structure of the code, and the main results, without including the heavy dataset and trained models.
The project explores a progression of unsupervised representation-learning techniques:
PCA → NMF → Autoencoder → Fine-Tuned DEC.
All experiments were conducted on a custom dataset of natural images created specifically for this assignment. The dataset is not included in this repository.
notebook/
- MG-unsupervised_final_project.ipynb
images/
docs/
- MG-unsupervised_learning_final_notebook.pdf
README.md
LICENSE
The notebook in this repository:
- preserves the original explanatory Markdown text,
- shows the full workflow conceptually,
- replaces all code outputs with static images,
- removes all training logic, datasets, models, and hyperparameters.
This ensures transparency for reviewers while preventing any reproduction or misuse of the original work.
To comply with the course rubric, respect the Honor Code and prevent academic plagiarism, the following items are intentionally excluded:
- dataset (images and metadata)
- trained model files and checkpoints
- full training scripts
- Android APK or assets
- any material enabling direct reuse of the project
This repository is intended exclusively for:
- course evaluation,
- peer review,
- and professional portfolio visibility.
It allows readers to understand the methodology and architecture, while protecting the integrity of the original implementation.
The original project used Python, PyTorch, scikit-learn, and common scientific libraries.
The dataset used in this project was manually generated for the assignment.
It is not included in this repository and is not required for peer review.
MIT License.