Hackatown 2020 - Second place in A.I. category
Mobile Application that detects certain type of product as recyclable or not.
Built with PyTorch, Java, Python Flask and MongoDB.
We achieved high accuracy prediction on selected items by using Transfer Learning
on the pre-trained ResNet50 network. We built our own dataset using
various preprocessing methods because there was none that we could find online.
McGill CodeJam 2018 Computer Vision Winner
Program that detects emotions based on facial expressions and calculate insurance fees for the user. The prediction was obtained by a 3 layer convolutional neural network
trained on 30,000 pictures for 15 epochs. Built with Python, with the help of OpenCV, Tensorflow and Keras. Won first place for Computer Vision category and third overall.
McGill MLH's Local Hack Day 2018
Application that uses artificial intelligence to help students study. We used machine learning to train a CNN for sleepiness predictions. Trained with 2 thousands pictures of people
with closed and opened eyes. Built with Python, with the help of OpenCV, Tensorflow and Keras. Won first place for Best Idea and Solution category during MLH's local hackday at McGill.