Deep Learning and Computer Vision Researcher
This section collects practical, implementation-oriented write-ups from my PhD research. I focus on robust computer vision and deep learning methods that transfer to real deployment settings: handling image quality variation, adapting models as data evolves, and improving segmentation and fine-grained classification under distribution shift.
Posts are written to be readable by both researchers and engineers. I include method framing, concrete lessons, failure cases, and links to the related papers.
A practical framework for preserving performance while adapting to evolving data distributions over time.
Lessons from no-reference quality scoring and cut-off point selection in FGVC pipelines.
A practical guide to using explainability maps to debug shortcut learning and improve confidence in model updates.