A research on anomaly detection models, Sep. 2020 – Feb. 2021

I was assigned to a project aimed at improving a pre-trained model that identifies pre-defined types of defects within images.

The project was divided into multiple smaller sections, each with its own objective. My role was to research, summarize, and coordinate state-of-the-art techniques and methodologies, making these advanced theories accessible to the model developers.

Several of the techniques we introduced were integrated into the model, leading to demonstrably better precision during the Proof-of-Concept (PoC) phase. Other methods, while not ultimately implemented, provided valuable insights that served as an instructive reference for different parts of the project.