Tri-path DINO: Feature Complementary Learning for Remote Sensing Multi-Class Change Detection PubID: pubid: 0000–0000/00$00.00 © 2021 IEEE
Kai Zheng, Hang-Cheng Dong, Shoulei Liu, Zhenkai Wu, Fupeng Wei, Lei Ding, Wei Zhang* · 2026
Read Tri-path DINO: Feature Complementary Learning for Remote Sensing Multi-Class Change Detection PubID: pubid: 0000–0000/00$00.00 © 2021 IEEE by Kai Zheng, Hang-Cheng Dong, Shoulei Liu, Zhenkai Wu, Fupeng Wei, Lei Ding, Wei Zhang* (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-2603-01498.txt.
Description
Kai Zheng, Hang-Cheng Dong, Shoulei Liu, and colleagues study how to spot many kinds of change between two satellite views. Their 2026 work targets fine detail, such as damage to single sites, where labels are scarce. Scenes vary a lot, which makes precise multi-class change maps hard to learn from scratch.
Their design, Tri-path DINO, adapts a large pre-trained vision model through three paths that learn together. One path reads broad scene traits, while a twin path builds fine detail step by step. A multi-scale attention step in the decoder then sharpens context at several sizes at once. The method tops tests on a Gaza damage data set and on a classic change set. Heat maps also show each path plays its planned role.
Catalog record
| CREATOR | Kai Zheng, Hang-Cheng Dong, Shoulei Liu, Zhenkai Wu, Fupeng Wei, Lei Ding, Wei Zhang* |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31368 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.01498