Capstone Projects

Object Detection: Boosted DETR

• Original model architecture combining the concept of model boosting with the existing DETR object detection model. This also includes an Tensorflow implementation of the original DETR (hybrid CNN / attention transformer) model, coded from first principles as presented in the paper End-to-End Object Detection with Transformers (2020) published by Facebook AI.

Image to Text Translation: Attention is What You Get

• Image CNN + Attention Encoder --> Attention + Dense text Decoder for the Bristol-Myers Squibb Molecular Translation challenge on Kaggle. Given a chemical diagram drawing, can we determine its specific molecule identification InChI code?
• Model based on the papers Attention is All You Need (2017) and Show, Attend and Tell: Neural Image Caption Generation with Visual Attention (2015)

Data Analysis and Classification: Predicting Water Reliability

• Capstone project from Wesleyan University's Data Analysis and Interpretation specialization on Coursera. Includes full written report (PDF) and Decision Tree-based predictive model.

RNN Encoder / Decoder: English -> German Translation

• Capstone project from Imperial College London's "Customizing Your Models With Tensorflow 2" class on coursera.

Neural Style Transfer: Web App

• Alter an image based on textural elements from a second image. (Web app implementation.)