Research process
This was my first research project, conducted during an internship at RoboLab at Prince Mohammad Bin Fahd University under Dr. Ghazanfar Latif. The end-to-end process covered the full ML research pipeline:
Data collection at local farms, image capture with a Nikon D3100, preprocessing with custom Python scripts for region-of-interest extraction, feature extraction and conventional ML experiments, followed by deep learning experiments — culminating in a published arXiv paper.
Pipeline
Field data collection
Visited local farms to gather real disease samples — no synthetic or web-scraped data.
Custom preprocessing
Built a Python tool to extract regions of interest from raw images, published separately on GitHub.
Classical ML baselines
Feature extraction and conventional ML experiments to establish baselines before deep learning.
Deep learning
CNN-based classification experiments, comparing architectures and training strategies on the custom dataset.
Built with
Python
TensorFlow
scikit-learn
OpenCV
NumPy
Matplotlib