GANimation
Project I
Project I
Re-implementation of GAN architecture conditioned on Action Units (AU) annotations. This model generates facial expressions in a continuous domain. This work is based on this paper and can control the intensity of a desired human expression.
This model works well under changing backgrounds and illumnination conditions. It is capable of changing expression of a person who has not been seen previously during the training. There are two main components: self-learned face attention which can focus into specific parts of the face and an anatomically aware expression representation which does not require a 3D face model nor the initialization method. Also, it's trained in an unsupervised manner.
Machine Learning (Found.)
Project II
Project II
The Boston housing market is highly competitive, and you want to be the best real estate agent in the area. To compete with your peers, you decide to leverage a few basic machine learning concepts to assist you and a client with finding the best selling price for their home. Luckily, you’ve come across the Boston Housing dataset which contains aggregated data on various features for houses in Greater Boston communities, including the median value of homes for each of those areas. Your task is to build an optimal model based on a statistical analysis with the tools available. This model will then be used to estimate the best selling price for your clients' homes.
Data: The modified Boston housing dataset consists of 489 data points, with each datapoint having 3 features. This dataset is a modified version of the Boston Housing dataset found on the UCI Machine Learning Repository. Features:
- RM: average number of rooms per dwelling
- LSTAT: percentage of population considered lower status
- PTRATIO: pupil-teacher ratio by town
- Target Variable 4. MEDV: median value of owner-occupied homes
Movie Recommendation System
Project III
Project III
An Implementation of Recommender Systems using aws, Django, python & Machine Learning algorithms such as Collaborative Filtering Algorithms (using Matrix Factorization and Neural Networks).
Implementation based upon: this DMF paper.
A matrix factorization model with a neural network architecture. Through the neural network architecture, users and items are projected into low dimensional vectors in a latent space.
Neighborhood Map
Project IV
Project IV
A single-page application featuring a map of your neighborhood.
Application utilizes Google's Map API and at least one additional third-party "data API". All data requests are retrieved in an asynchronous manner. In the event of a failed data retrieval errors are handled gracefully. Third-party API used: Wikipedia's API Additional functionality included in this application: map markers to identify popular locations or places you’d like to visit, a search function to easily discover these locations, and a listview to support simple browsing of all locations. Framework Used: Knockout Knockout is used to handle the list, filter, and any other information on the page that is subject to changing state. Things that are not be handled by Knockout: anything the Maps API is used for, creating markers, tracking click events on markers, making the map, refreshing the map. Note 1: Tracking click events on list items is handled with Knockout. Note 2: Creating your markers as a part of your ViewModel.
