yuhaozhang94/facial-frontalization yuhaozhang94/facial-frontalization

  • Stars
    star
    1
  • Language
    Python
  • Created over 4 years ago
  • Updated almost 4 years ago
Twitter logo Share LinkedIn logo Share Facebook logo Share
yuhaozhang94/facial-frontalization
github

yuhaozhang94

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Generative Adversarial Networks (GANs) have captured researchers’ interests in many domains, especially in facial image generation. This research project attempted to augment security efforts by generating a frontal image of a subject from its side profile captured by the camera using GANS.

More Repositories

1

humanpose-rnn-action-recognition

We implemented a deep neural network to predict human actions by modifying the gait recognition model by Marian Margeta (https://github.com/marian-margeta/gait-recognition). We extracted the spatial features of the human poses of each frame and extract their temporal features by running a Recurrent Neural Network. At last, we classify the human action with a Logistic Regression Model. We were able to achieve an accuracy of 94.87% on the G3D dataset (http://dipersec.kingston.ac.uk/G3D/G3D.html).
Jupyter Notebook
3
star
2

changi-airport-taxiway-planning

We want to solve the coupled routing and timing optimization problem in taxiway planning. There are many individual constraints such as push-back times, taxiway layouts, and separation rules all requiring considerations. The dynamic nature of this problem requires planning to be updated regularly. We propose to formulate the optimization problem in a Receding Horizon scheme, meaning that a plan is implemented in small execution horizons, before re-planning occurs. We will have Changi Airport surface modelled as a graph containing nodes, each representing a junction of taxiways. We will construct the connectivity and distance matrix between nodes. Active aircrafts are then modelled as points moving along the arcs, subject to speed limit. Each aircraft moves from its origin node to its destination node, while observing taxiway rules. The decision variables are therefore 1) the node at which an aircraft begins its kth move 2) the time at which an aircraft starts its kth move. The objective is to minimize a weighted combination of total taxi time of all aircrafts and their total taxi distance. We will use Mixed Integer Linear Programming and CPLEX to solve the optimization problem. As taxiway planning is coupled with runway scheduling and other operations, we made assumptions to simplify the problem in order to define a realistic scope given the time constraint.
Jupyter Notebook
2
star
3

song-recommender-system

Song recommendation was the subject of study in our group project. Using implicit ratings consisting of user ID, song ID and play count, we built a song recommender system with Weighted Matrix Factorization and provided explanations for the recommendation using both latent factors and additional song acoustic features. In dealing with the cold-start problem, we allowed the new user to either specify his/her artist preference or music acoustic preference, and recommended songs based on his/her nearest neighbours in the existing user base.
Jupyter Notebook
1
star