Danh Nguyen

Machine Learning Research Affiliate

• Research Background: Riot Games had a large system of champions and champion skins. The process of developing a new champion or skin required 3D modelling from concept art, which could require hundreds of hours of modelling. This research aimed to solve the problem by using machine learning to reconstruct 3D models from concept art to speed up the prototyping process.

• Worked together with data science experts from Riot Games and 2 student researchers to tackle challenge of 3D Humanoid Game Character Generation using machine learning with Python, PyTorch.

• Generated and pre-processed 2D game character dataset from 3D models to facilitate efficient visualisation, model training and evaluation.

• Contributed creative solution ideas and programmed state-of-the-art algorithms to create 3D models from concept art, reducing hundreds of hours of modelling.

Technologies:

Machine Learning Research Affiliate

• Research Background: In the development process of open-world games, Ubisoft requires the creation of a large number of artworks (even small details), which can be a great burden for artists. This research aimed to address one aspect of this challenge by using machine learning to reconstruct 3D animal models, animations.

• Cooperated with data scientists, artists from Ubisoft and 3 other researchers to resolve problem of 3D Animal Reconstruction from realistic and synthetic data.

• Restored 3D bird models, animations from images with Python, TensorFlow and PyTorch to help artists reduce workload in modelling detailed objects in game.

• Applied results to game prototype with Blender, Unreal Engine and C++ to ensure quality and performance of reconstructed models and animations.

Technologies:

Computer Vision Software Engineering Intern

• Constructed Computer Vision models for detection and classification problems with Python, Tensorflow and Pytorch, improving accuracy and speed of autonomous car perception system.

• Built an application for lightweight Traffic Sign real-time detection with Linux, C++ and CMake to enhance performance of perception component.

• Carried out Sensor Calibration, Data Processing and developed 3D Segmentation and Classification algorithms on LiDAR point cloud data.

Technologies:

Software Engineering Intern

• Researched mathematical morphology and manipulated Python, OpenCV, TensorFlow, Scikit-learn to build detection and classification models for book cover database.

• Integrated models in a smart library system application; designed, developed, tested, and documented new features for existing system.

• Pre-processed, stored book cover database, analysed data and machine learning model results with MySQL and Python.

Technologies: