Research
Face recognition under challenging conditions
Master’s research, Royal University of Phnom Penh (2024 – 2025). A comparative study of ResNet50 and FaceNet when faces are poorly lit, partly covered, or turned away.
83.44%
accuracy with ResNet50 under the evaluated settings
2
architectures compared: ResNet50 vs FaceNet
3
conditions: low light, occlusion, pose variation
Method
- Preprocessing
- Feature extraction
- Embeddings
- Similarity scoring
- Threshold evaluation
Images were preprocessed, then each model produced face embeddings. I compared identities using cosine similarity and Euclidean distance with tuned similarity thresholds, and evaluated performance separately for each challenging condition.
Why it matters to my design work
Research taught me to define a question, measure honestly, and report results with their limits. I bring the same habit to usability testing: decide what success looks like before looking at the data.
Earlier work
Other research & academic projects
2023
DataVue: Data Analytics Project
Korea Software HRD Center
Led a team building a system that scrapes online data and turns it into statistics and visualizations. Designed the end-to-end UX.
2022
Grant Proposal: Rural Primary School Project
Institute of Foreign Languages, RUPP
Proposal to support underprivileged students in rural Siem Reap with supplies, library resources, and hygiene facilities. Covered planning, budgeting, and impact evaluation.
2021
Parents’ Attitudes Toward Children’s Online Learning
Institute of Foreign Languages, RUPP
Co-authored qualitative study during COVID-19: interviews and thematic analysis of challenges in education, time management, parental support, and finances. Early user-research practice.