content-engine: publish ml-sharp card
Browse files
README.md
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
pretty_name: SHARP:单张图的秒级 3D 高斯视图合成
|
| 4 |
+
tags:
|
| 5 |
+
- 3D vision
|
| 6 |
+
- gaussian splatting
|
| 7 |
+
- view synthesis
|
| 8 |
+
- Apple
|
| 9 |
+
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# SHARP:单张图的秒级 3D 高斯视图合成
|
| 13 |
+
|
| 14 |
+
SHARP 是 Apple 提出的单目视图合成方法:给定一张照片,它在单次前馈中回归出场景的 3D 高斯表示,标准 GPU 上耗时不到 1 秒。得到的 3DGS 可以被实时渲染,输出附近视角的高分辨率写实图像,而且表示是 metric 的,带绝对尺度,支持米制相机运动。 按 README 与论文摘要,它在多个数据集上把 LPIPS 降低 25–34%、DISTS 降低 21–43%,同时把合成耗时降低三个数量级。 对开发者来说,值得看的是它把「重建」压缩成了一次前馈调用,并且给出了现成的 CLI 与权重。
|
| 15 |
+
|
| 16 |
+
This repository collects reading notes and links. **No model weights are redistributed here** — the
|
| 17 |
+
official checkpoint and code stay with their authors; follow the links below.
|
| 18 |
+
|
| 19 |
+
## Key facts
|
| 20 |
+
- **合成耗时**: 标准 GPU 上不到 1 秒,单次前馈
|
| 21 |
+
- **LPIPS**: 较此前最佳模型降低 25–34%
|
| 22 |
+
- **DISTS**: 较此前最佳模型降低 21–43%
|
| 23 |
+
- **加速幅度**: 合成耗时降低三个数量级
|
| 24 |
+
- **输出表示**: metric 3D 高斯(3DGS),输出 .ply 文件
|
| 25 |
+
- **泛化**: 论文称具备鲁棒的零样本泛化
|
| 26 |
+
|
| 27 |
+
## Sources
|
| 28 |
+
- SHARP 通过单次前馈在标准 GPU 上以不到 1 秒的时间回归出场景的 3D 高斯表示参数。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 29 |
+
- SHARP 产出的 3D 高斯表示可以被实时渲染,生成附近视角的高分辨率写实图像。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 30 |
+
- 该表示是 metric 的,带绝对尺度,支持米制的相机运动。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 31 |
+
- 实验结果展示了 SHARP 跨数据集的鲁棒零样本泛化能力。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 32 |
+
- 在多个数据集上,SHARP 相对此前最佳模型把 LPIPS 降低 25–34%、DISTS 降低 21–43%,同时把合成耗时降低三个数量级。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 33 |
+
- 预测结果是以 3D 高斯泼溅(3DGS)形式存放在输出目录中的可渲染结果。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 34 |
+
- 官方 CLI 的预测入口是 sharp predict,用 -i 指定输入图像目录、-o 指定输出高斯目录。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 35 |
+
- 模型权重在第一次运行时自动下载,并缓存到 ~/.cache/torch/hub/checkpoints/。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 36 |
+
- 高斯预测在 CPU、CUDA 和 MPS 上都能工作,但通过 --render 渲染相机轨迹视频目前需要 CUDA GPU。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 37 |
+
- 输出的 3DGS 遵循 OpenCV 坐标约定(x 右、y 下、z 前),场景中心大致在 (0, 0, +z),接第三方渲染器时需要自行缩放与旋转来重新居中。 — [raw.githubusercontent.com](https://raw.githubusercontent.com/apple/ml-sharp/main/README.md)
|
| 38 |
+
|
| 39 |
+
## Links
|
| 40 |
+
- [arXiv 2512.10685](https://arxiv.org/abs/2512.10685)
|
| 41 |
+
- [项目主页](https://apple.github.io/ml-sharp/)
|
| 42 |
+
- [官方仓库](https://github.com/apple/ml-sharp)
|
| 43 |
+
- [模型权重](https://ml-site.cdn-apple.com/models/sharp/sharp_2572gikvuh.pt)
|
| 44 |
+
|
| 45 |
+
## More
|
| 46 |
+
|
| 47 |
+
SHARP 展示了一件值得记住的事:一张普通照片里,本来就藏着足够多的信息,只要有一个足够鲁棒的表示方式把它读出来。 在你自己的照片上,最实用的第一步往往不是重建 3D,而是先把图像本身的质量拉起来。 DLSS 5 是一个在线 AI 图像编辑器,把手里已有的图画上传上去,就能试试我们的 AI 超分与画质增强。
|
| 48 |
+
|
| 49 |
+
[上传图片,试试 AI 增强](https://www.dlss5nvidia.com)
|
| 50 |
+
|
| 51 |
+
---
|
| 52 |
+
|
| 53 |
+
Notes compiled from `topics/ml-sharp/` with content-engine.
|