Stronger Semantic Encoders Can Harm Relighting Performance: Probing Visual Priors via Augmented Latent Intrinsics
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910007826579456 |
|---|---|
| author | Xing, Xiaoyan Zhang, Xiao Karaoglu, Sezer Gevers, Theo Bhattad, Anand |
| author_facet | Xing, Xiaoyan Zhang, Xiao Karaoglu, Sezer Gevers, Theo Bhattad, Anand |
| contents | Image-to-image relighting requires representations that disentangle scene properties from illumination. Recent methods rely on latent intrinsic representations but remain under-constrained and often fail on challenging materials such as metal and glass. A natural hypothesis is that stronger pretrained visual priors should resolve these failures. We find the opposite: features from top-performing semantic encoders often degrade relighting quality, revealing a fundamental trade-off between semantic abstraction and photometric fidelity. We study this trade-off and introduce Augmented Latent Intrinsics (ALI), which balances semantic context and dense photometric structure by fusing features from a pixel-aligned visual encoder into a latent-intrinsic framework, together with a self-supervised refinement strategy to mitigate the scarcity of paired real-world data. Trained only on unlabeled real-world image pairs and paired with a dense, pixel-aligned visual prior, ALI achieves strong improvements in relighting, with the largest gains on complex, specular materials. Project page: https:\\augmented-latent-intrinsics.github.io |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01391 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Stronger Semantic Encoders Can Harm Relighting Performance: Probing Visual Priors via Augmented Latent Intrinsics Xing, Xiaoyan Zhang, Xiao Karaoglu, Sezer Gevers, Theo Bhattad, Anand Computer Vision and Pattern Recognition Image-to-image relighting requires representations that disentangle scene properties from illumination. Recent methods rely on latent intrinsic representations but remain under-constrained and often fail on challenging materials such as metal and glass. A natural hypothesis is that stronger pretrained visual priors should resolve these failures. We find the opposite: features from top-performing semantic encoders often degrade relighting quality, revealing a fundamental trade-off between semantic abstraction and photometric fidelity. We study this trade-off and introduce Augmented Latent Intrinsics (ALI), which balances semantic context and dense photometric structure by fusing features from a pixel-aligned visual encoder into a latent-intrinsic framework, together with a self-supervised refinement strategy to mitigate the scarcity of paired real-world data. Trained only on unlabeled real-world image pairs and paired with a dense, pixel-aligned visual prior, ALI achieves strong improvements in relighting, with the largest gains on complex, specular materials. Project page: https:\\augmented-latent-intrinsics.github.io |
| title | Stronger Semantic Encoders Can Harm Relighting Performance: Probing Visual Priors via Augmented Latent Intrinsics |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.01391 |