Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.19964 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914414233387008 |
|---|---|
| author | Zhang, Tianbao Liang, Zhenyu Song, Zhenbo Wang, Nana Zhang, Xiaomei Cai, Xudong Zhu, Zheng Wu, Kejian Wang, Gang Fan, Zhaoxin |
| author_facet | Zhang, Tianbao Liang, Zhenyu Song, Zhenbo Wang, Nana Zhang, Xiaomei Cai, Xudong Zhu, Zheng Wu, Kejian Wang, Gang Fan, Zhaoxin |
| contents | High-resolution geometric prediction is essential for robust perception in autonomous driving, robotics, and AR/MR, but current foundation models are fundamentally limited by their scalability to real-world, high-resolution scenarios. Direct inference on 2K images with these models incurs prohibitive computational and memory demands, making practical deployment challenging. To tackle the issue, we present 2K Retrofit, a novel framework that enables efficient 2K-resolution inference for any geometric foundation model, without modifying or retraining the backbone. Our approach leverages fast coarse predictions and an entropy-based sparse refinement to selectively enhance high-uncertainty regions, achieving precise and high-fidelity 2K outputs with minimal overhead. Extensive experiments on widely used benchmark demonstrate that 2K Retrofit consistently achieves state-of-the-art accuracy and speed, bridging the gap between research advances and scalable deployment in high-resolution 3D vision applications. Code will be released upon acceptance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_19964 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | 2K Retrofit: Entropy-Guided Efficient Sparse Refinement for High-Resolution 3D Geometry Prediction Zhang, Tianbao Liang, Zhenyu Song, Zhenbo Wang, Nana Zhang, Xiaomei Cai, Xudong Zhu, Zheng Wu, Kejian Wang, Gang Fan, Zhaoxin Computer Vision and Pattern Recognition High-resolution geometric prediction is essential for robust perception in autonomous driving, robotics, and AR/MR, but current foundation models are fundamentally limited by their scalability to real-world, high-resolution scenarios. Direct inference on 2K images with these models incurs prohibitive computational and memory demands, making practical deployment challenging. To tackle the issue, we present 2K Retrofit, a novel framework that enables efficient 2K-resolution inference for any geometric foundation model, without modifying or retraining the backbone. Our approach leverages fast coarse predictions and an entropy-based sparse refinement to selectively enhance high-uncertainty regions, achieving precise and high-fidelity 2K outputs with minimal overhead. Extensive experiments on widely used benchmark demonstrate that 2K Retrofit consistently achieves state-of-the-art accuracy and speed, bridging the gap between research advances and scalable deployment in high-resolution 3D vision applications. Code will be released upon acceptance. |
| title | 2K Retrofit: Entropy-Guided Efficient Sparse Refinement for High-Resolution 3D Geometry Prediction |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.19964 |