The authors built ARROW around an order-invariant image-set encoder derived from Depth Anything 3.
Content-based frame identities unify 3D reconstruction and point tracking
ARROW processes unordered, unposed image sets and answers queries about a point’s 3D position across views and capture times.
Academic
Ilya Fradlin · Christian Schmidt · Jens Piekenbrinck · Karim Knaebel · Gonzalo Martin Garcia · Bastian Leibe
RWTH Aachen University
Research Digest··2 min read
Fradlin et al.
Why this paper
From RWTH Aachen University
In one line
ARROW unifies 3D reconstruction and point tracking across unordered images and videos, achieving state-of-the-art 3D tracking while remaining competitive on reconstruction.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§