ECSE 415: Intro to Computer Vision - Final Project

This repository contains our classifier and localization implementations for the ECSE 415 final project. The goal of this project was to implement an SVM classifier and a localization method. Deep-learning based methods for both localization and classification were also accepted.

Classifier

• SVM Classifer
• CNN Classifier

Localization

Custom-trained version of YOLOv3 using darknet.

Authors

• Tristan Toupin
• Matthew Lesko
• Alexander Harris
• Abbas Yadollahi
• Filip Bernevec

Dataset

The datasets used were provided by the following paper:

Z. Luo, F.B.Charron, C.Lemaire, J.Konrad, S.Li, A.Mishra, A. Achkar, J. Eichel, P-M Jodoin,
“MIO-TCD: A new benchmark dataset for vehicle classification and localization
in press at IEEE Transactions on Image Processing”, 2018.

Link to dataset: http://podoce.dinf.usherbrooke.ca/challenge/dataset/