Image classification with transfer learning, classical feature extraction and filtering, and a KNN classifier implemented from scratch
Computer vision spanning both ends of the field: hand-rolled classical techniques and modern deep transfer learning.
The classical side implements image filters and feature extraction directly, plus a K-Nearest-Neighbours image classifier written from scratch rather than imported, to work through the mechanics rather than the API.
The deep learning side applies CNNs and transfer learning to two real classification problems: plant species identification and sports product categorisation.
code/ source code, notebooks and datasets
docs/ reports, papers, slide decks and recordings
Feature_Extraction.ipynb-- 306 KBFilters.ipynb-- 8.9 MBKNN_From_Scratch.ipynb-- 1.2 MBPlants_Classification.html-- 6.4 MBSport_Products_Classification.ipynb-- 7.4 MB
Sport_Products_Classification.pdf-- 1.9 MBSport_Products_Classification.pptx-- 20.1 MB
computer-vision · image-classification · opencv · transfer-learning · cnn · feature-extraction · deep-learning
Part of my AI and Data Science portfolio. Work produced during my M.Eng. in AI and Data Science at the University of Ottawa and my B.Sc. in Computer Science.