UIP2P: Unsupervised Instruction-based Image Editing via Edit Reversibility Constraint
Enis Simsar, Alessio Tonioni, Yongqin Xian, Thomas Hofmann, Federico Tombari · 2024
Read UIP2P: Unsupervised Instruction-based Image Editing via Edit Reversibility Constraint by Enis Simsar, Alessio Tonioni, Yongqin Xian, Thomas Hofmann, Federico Tombari (2024) 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-2412-15216.txt.
Description
Enis Simsar and co-authors present UIP2P in a 2024 paper on instruction based image editing without supervision. Existing methods need triples of input image, edited image, and instruction. Building those triples with other editors adds bias, while human labeling is costly and limits generalization.
Their method applies a forward edit and a reverse edit in one training step. A reversibility constraint aligns image, text, and attention between the two directions. That removes the need for ground truth edited images. Training can use plain image caption pairs, and tests showed strong fidelity across a broad range of edits.
Catalog record
| CREATOR | Enis Simsar, Alessio Tonioni, Yongqin Xian, Thomas Hofmann, Federico Tombari |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 63 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-15216