Aditya Nagachandra

engineer

Publications

VLM4D: Towards Spatiotemporal Awareness in Vision‑Language Models
ICCV 2025 · Research
Authors
Shijie Zhou*, Alexander Vilesov*, Xuehai He*, Ziyu Wan, Shuwang Zhang, Aditya Nagachandra, Di Chang, Dongdong Chen, Xin Eric Wang, Achuta Kadambi
UCLA · Microsoft · UCSC · USC
* Equal contribution

Projects

What I learned today

April 6, 2026
3DGS initialization importance and mesh accuracy metrics
April 5, 2026
How a CO2 sensor works
April 4, 2026
Multi-view diffusion model conditioning with Plückner coordinates
April 3, 2026
Quaternion greedy traversal — WorldMesh camera layout

WorldMesh orders multi-view captures with a greedy nearest-neighbor walk using quaternion similarity |q1·q2| between consecutive cameras (2 bootstrap + 16 perimeter + 8 overhead). Below is a small interactive top-down sketch of that 2+16+8 layout and the traversal edges.

Quaternion greedy traversal — WorldMesh 2+16+8 camera layout c1 short c0 short long wall long wall Bootstrap c0, c1 Perimeter ×16 Overhead ×8 Traversal edge press play — greedy nearest-neighbor by |q1·q2| 0 / 26
0 / 26