Case Study: American University of Beirut achieves 46% faster image labeling for dietary ML research with Labelbox

American University of Beirut takes on dietary behavior research with ML

American University of Beirut researchers, funded by the International Development Research Centre (IDRC), set out to build two convolutional neural network models to classify wearable‑camera images for dietary behavior research (e.g., identifying when the camera‑wearer or others are consuming food, food outlets, or food ads). Their challenge was obtaining high‑quality labeled data for supervised deep learning and reducing the time and cost of traditional, labor‑intensive dietary analysis, so they engaged Labelbox for annotation support.

Labelbox provided an intuitive annotating interface, QA tools, a clear raw‑data‑to‑production pipeline, and Labelbox Workforce labeling operations to speed labeling and ensure quality. Using Labelbox, the team cut labeling time from 13 seconds per image to 7 seconds per image (a 46% reduction), lowered costs, and accelerated model training and scalability, enabling faster, more reliable dietary analysis.

American University of Beirut

Zoulfikar Shmayssani

Researcher