Machine Learning Projects

A curated collection of machine learning projects covering various domains including computer vision, predictive modeling, and data analysis.

Projects

Project	Description
Digit Recognition	Streamlit web app for handwritten digit recognition using CNN (MNIST)
Digiti Signum	Fingerprint matching & alteration detection using classical CV and deep learning
Understanding Data Augmentation	ResNet18 CIFAR-10 experiment comparing models with/without augmentation
E-commerce Conversion Prediction	AI model predicting visitor purchase conversion for LushAura
Heat Pump Contribution Analysis	Deep learning and analyzing heat pump energy contribution using real data
Can't See the Forest...	Univariate analysis + decision trees/random forests for student performance
Data Processing	Data preprocessing utilities + Chicago crime data analysis
CNNs from Scratch	CNN architectures (EfficientNet/Inception-inspired) on CIFAR-10

Notable Results

• 99.47% fingerprint alteration detection (ResNet18 transfer learning).

• 95.17% CIFAR-10 accuracy with data augmentation (+6.4% improvement over baseline).

• 99% handwritten digit recognition (MNIST CNN).

• 98.40% multi-class fingerprint classification (4 alteration types).

Technologies

• Deep Learning: PyTorch, TensorFlow/Keras.

• Machine Learning: scikit-learn.

• Computer Vision: OpenCV, scikit-image.

• Data Processing: pandas, numpy.

• Visualization: matplotlib, seaborn.

• Web Apps: Streamlit.

• Notebooks: Jupyter.

Getting Started

Clone this repository with all submodules:

git clone --recurse-submodules https://github.com/HelixCipher/machine-learning.git

To update all submodules to the latest version:

git submodule update --remote

License

Each sub-project is licensed under its own terms. See individual project repositories for details.