SMFormer: Empowering Self-supervised Stereo Matching via Foundation Models and Data Augmentation
Yun Wang, Zhengjie Yang*, Jiahao Zheng, Zhanjie Zhang, Dapeng Oliver Wu*, Yulan Guo · 2026
Read SMFormer: Empowering Self-supervised Stereo Matching via Foundation Models and Data Augmentation by Yun Wang, Zhengjie Yang*, Jiahao Zheng, Zhanjie Zhang, Dapeng Oliver Wu*, Yulan Guo (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-10218.txt.
Description
Wang and five colleagues improve self-taught stereo matching with SMFormer, guided by a base model and image tweaks. Self-taught methods assume matching points look alike across views, but real noise breaks that match and leaves invalid guidance. SMFormer pairs the base model with a feature pyramid network for robust cues. Its training tweaks force a match between learned and shifted cues, and steady the match between tweaked and standard depth guesses. SMFormer leads self-taught methods on benchmarks and rivals labeled ones, even beating some on Booster.
Catalog record
| CREATOR | Yun Wang, Zhengjie Yang*, Jiahao Zheng, Zhanjie Zhang, Dapeng Oliver Wu*, Yulan Guo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59326 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.10218