In this five week project, was funded to develop a web application for the Zoological Society of London to train machine learning models to automate ecological data collection. Features include dataset creation, annotation, train/validation and test set splitting, training, and inference.

The tech stack used is React for front-end, and Express and Flask for back-end (Flask used to create a REST API to enable submitting of training/inference jobs to HPC).

Dataset Management

Annotator

Dataset Split Creation

Model Training

Training Monitoring

Inference and Population Data Analysis