Development of a 2D video game | 2019 - Current            

  • Led a team of 3 to design and develop a video game from scratch in Unity environment by using C# to code the gameplay mechanics and building a sleek UI for players. 

  • Implemented parallax effect using C# to create the illusion of distance between 4+ layers.

  • Import prebuilt assets into current projects and designed 5 new robust and fun levels.  


Art Generation with neural style transfer | 2020                                                               

  • Utilized pretrained VGG-19 weights to implement a model to minimize content and style cost function by taking activation from specific layers in the architecture and using adam optimization algorithm in tensorflow.

  • Generated art using the style from one image to create content similar to “content” image using said model.

Face Verification and Recognition | 2020 

  • Inspired by FaceNet, implemented triplet loss function in keras and tensorflow

  • Used pretrained weights from FaceNet to encode facial images into 128-dimensional vectors 

  • Utilized said encodings to perform face verification and recognition by computing the L2 distance between input encodings and database encodings.


Autonomous Driving Application Car Detection | 2020     

  • Implemented a filtering function to filter “boxes” of classes below a threshold.

  • Utilized pretrained YOLOv2 model parameters and implemented non max suppression and IoU to reduce multiple detections of same objects in an image

Movie Recommender System | 2020                 

  • Designed and implemented a movie recommender system using collaborative filtering algorithm from movielens 100K dataset.

  • The system recommends a set of movies with predicted ratings based on users previous rating history.




Alpha capture of playable character and an enemy sprite. 

© 2020 by Akhil Thalatoti.