🎓 Free Capstone Projects with Full Documentation, ER Diagrams & Source Code — Updated Weekly for 2026
👨‍💻 Free Source Code & Capstone Projects for Developers

Frequently Asked Questions

Are these ML projects free for capstone and thesis use?
Yes. All projects are free to download, modify, and submit. No attribution required for academic use. Most are MIT-licensed. The underlying libraries (OpenCV, MediaPipe, TensorFlow, scikit-learn) are open source and free for any purpose.
Do I need a GPU for these projects?
No. Every project here runs on CPU. The face/hand/pose models use MediaPipe — designed for real-time CPU inference. The handwritten digit recognition CNN trains on CPU in under 5 minutes with MNIST. If you want to add a custom-trained model later, use Google Colab's free GPU for training and ship CPU inference in your capstone.
What is the difference between Machine Learning, Deep Learning, and Computer Vision?
Machine Learning is the broad field; algorithms that learn patterns from data (decision trees, k-means, SVM, neural networks). Deep Learning is a subset of ML using multi-layer neural networks (CNNs, Transformers). Computer Vision is the application domain (images and video); typically uses ML or deep learning under the hood. Most projects here are CV applications using pretrained deep-learning models. We have a separate Deep Learning Projects category for thesis-grade DL work.
Can I pass an ML/AI project as a full capstone?
Yes — if you wrap the core ML capability in a real-world application with users, data persistence, and reports. Examples: Face Recognition Attendance System (face detection plus database plus reports), Drowsiness Alert for Drivers (eye-blink detection plus audio alert plus log dashboard). Standalone ML demos without a domain shell are too thin for capstone scope.
OpenCV vs MediaPipe vs TensorFlow — when to use each?
OpenCV for video I/O, image processing (filters, edges, blending), and Haar-cascade detection. MediaPipe for state-of-the-art face/hand/pose landmark detection — designed for real-time CPU. TensorFlow / PyTorch for training custom models (image classification, custom object detection). Most projects here use OpenCV plus MediaPipe together — install: pip install opencv-python mediapipe.
What if Python 3.13 breaks OpenCV / MediaPipe?
Downgrade to Python 3.11 or 3.12 — stable wheels available for both. Python 3.13 dropped distutils and some ML library wheels lag behind. Use pyenv (macOS/Linux) or pyenv-win to manage multiple Python versions side-by-side.
How often is this ML projects list updated?
New ML/AI projects are added regularly. Last refreshed May 2026. 2026 additions: YOLO v8/v9 object detection, MediaPipe Tasks API, LangChain RAG demos.