Caner Hazirbas

Research Scientist  ·  Meta AI  ·  On-Device ML & GenAI

I am a Research Scientist at Meta AI working on on-device ML and GenAI for photo and video edits. Before that I was a Deep Learning Engineer at Apple, and completed my PhD at TU Munich in the group of Daniel Cremers. My research interests span depth estimation, RGB-D scene understanding, visual representation learning, responsible AI, and on-device machine learning.

Datasets

Responsible AI

Casual Conversations v1

Meta AI · 2021

A dataset of 45,186 videos featuring paid participants to help researchers evaluate AI models for fairness across age, gender, and skin tone attributes.

Responsible AI

Casual Conversations v2

Meta AI · 2023

An expanded dataset with 26,467 videos covering additional fairness dimensions including perceived age, gender, physical attributes, and self-reported annotations for more robust AI evaluation.

Focal Stack

DDFF 12-Scene

Benchmark · 2017

Large-scale focal stack benchmark with 12 indoor scenes and dense depth ground truth — 25× larger than prior benchmarks, enabling ML-based depth-from-focus research.

RGB-D

TUM-LSI

TU Munich

Large-scale indoor RGB-D dataset for scene understanding and semantic segmentation research, captured across real-world environments at TU Munich.

Publications All on Scholar

News