Scholar
Julian Straub
Google Scholar ID: 49_cCT8AAAAJ
Research Scientist, Reality Labs Research
Citations & Impact
Citations
9,419
H-index
20
i10-index
28
Publications
20
Co-authors
95
list available
Publications
Cited
0
Resume (English only)
Academic Achievements
- CVPR 2025: Published 'Sonata: Self-Supervised Learning of Reliable Point Representations', achieving 3x higher linear probing accuracy on ScanNet and 2x performance with only 1% training data.
- Arxiv 2024: Led the release of 'EFM3D: A Benchmark for Measuring Progress Towards 3D Egocentric Foundation Models', establishing the first benchmark and baseline model EVL.
- CVPR 2024: Co-authored 'EgoLifter: Open-world 3D Segmentation for Egocentric Perception', using Gaussian Splats for reconstruction and instance segmentation of Project Aria data.
- CVPR 2023: Contributed to Omni3D, a large-scale 3D object detection benchmark with 234k images, 3M+ instances, and 98 categories; proposed the generalizable Cube R-CNN model.
- CVPR 2023: Co-developed OrienterNet, the first deep network to localize images using 2D semantic maps from OpenStreetMap.
- CVPR 2023: Co-proposed PARQ for multi-view 3D object detection from short posed video sequences.
- Invited speaker at major conferences including ICCV 2025, CVPR 2025, and ECCV 2024 on topics like 3D perception, point clouds, and egocentric AI.
