pywavelets

Image fusion using Discrete Wavelet Transformation

In this repo I am trying to build test automation framework using machine learning in the error detection and prediction. By using ML the framework will be able to recover from the issue will face during execution process.

Python helper for computing and visualising the Continuous Wavelet Transform (CWT) on 1-D signals such as audio waveforms. And for computing the Haar Wavelet Transform in regards to Neural Network Compression and inference after compression. Powered by NumPy · SciPy · PyWavelets · Matplotlib · Torch · Transformers.

This website helps to classifies celebrities namely actors or actresses based on their images. This model classifies 5 celebrities only. If the image is not clear, the model doesn't classify the images. However, if a match is found, the name of the celebrity it matches is mentioned.

Python implementation and benchmark of classical ECG denoising methods — wavelet (hard/soft/improved), translation-invariant, FFT, LMS, and Wiener-Hopf — compared by SNR.

4. Transformer-XH: Hybrid Attention for Long Horizons (PyTorch with PyWavelets). Hibridne arhitekture za predikciju koje kombinuju deep learning i klasične time-series modele.

Implemented and compared three deep learning architectures - ANN, 1D-CNN, and LSTM with a preprocessing pipeline including DWT denoising and SMOTE balancing across 1M+ samples. The best model (LSTM) achieved 99.73% accuracy and 99.70% macro F1, with all experiments tracked via MLflow.

Machine Learning algorithms implemented from scratch

Facial Recognition with help of Haar Cascade files

An end-to-end Machine Learning pipeline and Flask web application built to recognize global sports icons. The system leverages OpenCV Haar Cascades for facial detection, Wavelet Transforms for feature extraction, and a GridSearch-optimized SVM model to deliver highly accurate predictions through a responsive, modern UI.

Laboratory projects created at WUST or related to my field of study

This repository contains my solutions for homeworks and exercises of Computer Vision course at the Ferdowsi University of Mashhad, Fall 2022

A machine learning model where we classify famous Nigerian politicians. We restrict classification to only 4 people

Trinetra is a powerful AI-driven application that helps users prepare, clean, and enhance various types of data including CSV files, PDFs, and images.

Time Series Analysis project for BDA at WUST containing CWT on dataset and NN training performed on transformed data

Signal Feature Learning

A Python-based quantitative engine utilizing 1D Discrete Wavelet Transforms (DWT) to filter stochastic noise from high-frequency financial tick data. Repurposed from advanced 2D academic image processing algorithms to demonstrate mathematical translation and modular software architecture.

Ballistocardiography (BCG) signal processing toolbox: preprocessing, spectral analysis, wavelet, baseline removal.