Projects
Engineering and research work: deep learning, signal processing and image processing, with measured results.
Research and machine learning
Deep Learning · Self-Supervised LearningDistill-SSLSelf-Supervised Teachers, Small StudentsA 25-experiment study on STL-10: pretrain ResNet-18 teachers with SimCLR, BYOL, and DINO, then distill them into a MobileNetV3-Small student with single, ensemble, and progressive strategies.Deep Learning · Medical ImagingMRI Brain Tumor SegmentationThree U-Net variants compared for brain tumor segmentation on 2D slices of BraTS-Africa MRI (four modalities), with a Streamlit app to compare predictions with the ground truth.Deep Learning · Signal ProcessingBearing Fault Diagnosis with a 1D CNNA 1D CNN that classifies bearing faults from raw vibration signals (CWRU dataset), tested on an unseen motor load and under added noise, with a Streamlit app for uploading a signal.
Smaller projects
Image Processing · Computer VisionDIP-LibInteractive Digital Image Processing ToolkitA Streamlit app for learning classical image processing: nine modules with live parameters, stacked into a pipeline that shows the image after every step.Signal Processing · Image ProcessingFrequency Decomposition ExplorerAn interactive Streamlit app to visualize, isolate, and reconstruct 1D or 2D signals based on their frequency bands using FFT and IFFT.Deep Learning · Time SeriesWind Power ForecastingForecasting a wind turbine's power output: a direct LSTM against a two-stage LSTM → Random Forest pipeline. The direct model halves the error.