Skip to content

GitLab

Commit d95cd255 authored by fengzch-das's avatar fengzch-das
Browse files

update readme

parent 79d832ca
Pipeline #2867 failed with stages
in 0 seconds
Show whitespace changes
Inline Side-by-side
Showing with 151 additions and 73 deletions
README.md
View file @ d95cd255
# DRTK – Differentiable Rendering Toolkit
This package is a PyTorch library that provides functionality for differentiable rasterization.
It consists of five main components:
* **transform**
* **rasterize**
* **render**
* **interpolate**
* **edge_grad**
There are also optional components such as **msi** and **mipmap_grid_sampler**. New components may be added in the future.
A typical flow looks like this:
**transform** → **rasterize** → **render** → **interpolate** → **CUSTOM SHADING** → **edge_grad**
where:
- **transform**: projects the vertex positions from camera space to image space
- **rasterize**: performs rasterization, where pixels in the output image are associated with triangles
- **render**: computes depth and baricentric image
- **interpolate**: interpolates arbitrary vertex attributes
- **CUSTOM SHADING**: user implemented shading
- **edge_grad**: special module that computes gradients for the **rasterize** step, which is not differentiable on its own. For details, please see [**Rasterized Edge Gradients: Handling Discontinuities Differentiably**](https://arxiv.org/abs/2405.02508)
## Hello Triangle
The "Hello Triangle" with DRTK would look like this:
```python
import drtk
import torch as th
from torchvision.utils import save_image # to save images
# create vertex buffer of shape [1 x n_vertices x 3], here for triangle `n_vertices` == 3
v = th.as_tensor([[[0, 511, 1], [255, 0, 1], [511, 511, 1]]]).float().cuda()
# create index buffer
vi = th.as_tensor([[0, 1, 2]]).int().cuda()
# rasterize
index_img = drtk.rasterize(v, vi, height=512, width=512)
# <div align="center"><strong>DRTK – Differentiable Rendering Toolkit</strong></div>
## 简介
这个软件包是一个 PyTorch 库，提供了 可微分栅格化 的功能.
## 安装
组件支持组合
| PyTorch版本 | fastpt版本 |DRTK版本 | DTK版本 | Python版本 | 推荐编译方式 |
| ----------- | ----------- | ----------- | ------------------------ | -----------------| ------------ |
| 2.5.1 | 2.1.0 |main-3c14f46 | >= 25.04 | 3.8、3.10、3.11 | fastpt不转码 |
| 2.4.1 | 2.0.1 |main-3c14f46 | >= 25.04 | 3.8、3.10、3.11 | fastpt不转码 |
| 其他 | 其他 | 其他 | 其他 | 3.8、3.10、3.11 | hip转码 |
+ pytorch版本大于2.4.1 && dtk版本大于25.04 推荐使用fastpt不转码编译。
### 1、使用pip方式安装
drtk whl包下载目录：[光和开发者社区](https://download.sourcefind.cn:65024/4/main)，选择对应的pytorch版本和python版本下载对应drtk的whl包
```shell
pip install torch* (下载torch的whl包)
pip install fastpt* --no-deps (下载fastpt的whl包)
source /usr/local/bin/fastpt -E
pip install drtk* (下载的drtk的whl包)
```
### 2、使用源码编译方式安装
# compute baricentrics
_, bary = drtk.render(v, vi, index_img)
#### 编译环境准备
提供基于fastpt不转码编译：
# we won't do shading, we'll just save the baricentrics and filter out the empty region
# which is marked with `-1` in `index_img`
img = bary * (index_img != -1)
1. 基于光源pytorch基础镜像环境：镜像下载地址：[光合开发者社区](https://sourcefind.cn/#/image/dcu/pytorch)，根据pytorch、python、dtk及系统下载对应的镜像版本。
save_image(img, "render.png")
2. 基于现有python环境：安装pytorch，fastpt whl包下载目录：[光合开发者社区](https://sourcefind.cn/#/image/dcu/pytorch)，根据python、dtk版本,下载对应pytorch的whl包。安装命令如下：
```shell
pip install torch* (下载torch的whl包)
pip install fastpt* --no-deps (下载fastpt的whl包, 安装顺序，先安装torch，后安装fastpt)
pip install pytest
pip install wheel
```
![hello triangle](docs/source/_static/hellow_triangle.png)
## Dependencies
* PyTorch >= 2.1.0
#### 源码编译安装
- 代码下载
```shell
git clone http://developer.sourcefind.cn/codes/OpenDAS/drtk.git # 根据编译需要切换分支
```
- 提供2种源码编译方式（进入drtk目录）：
```
1. 设置不转码编译环境变量
source /usr/local/bin/fastpt -C
## Installing
2. 编译whl包并安装
python3 setup.py build_ext bdist_wheel
pip install dist/drtk* --no-deps
To build a wheel and install it:
```
pip install git+https://github.com/facebookresearch/DRTK.git
3. 源码编译安装
python3 setup.py build_ext install --no-deps
```
#### 注意事项
+ 若使用pip install下载安装过慢，可添加pypi清华源：-i https://pypi.tuna.tsinghua.edu.cn/simple/
+ ROCM_PATH为dtk的路径，默认为/opt/dtk
+ 在pytorch2.5.1环境下编译需要支持c++17语法，打开setup.py文件，把文件中的 -std=c++14 修改为 -std=c++17
To build inplace, which is useful for package development:
## 验证
```
python setup.py build_ext --inplace -j 1
python3
Python 3.10.12 (main, Feb 4 2025, 14:57:36) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import drtk
>>> drtk.__version__
'0.1.0'
>>>
```
版本号与官方版本同步，查询该软件的版本号，例如0.3；
## Contributing
See the [CONTRIBUTING](CONTRIBUTING.md) file for how to help out.
## Known Issue
- 无
## License
DRTK is MIT licensed, as found in the [LICENSE](LICENSE) file.
## Citation
```bibtex
@mish{pidhorskyi2024rasterized,
title = {Rasterized Edge Gradients: Handling Discontinuities Differentiably},
author = {Pidhorskyi, Stanislav and Simon, Tomas and Schwartz, Gabriel and Wen, He and Sheikh, Yaser and Saragih, Jason},
howpublished = {arXiv preprint},
year = {2024},
note = {arXiv:2405.02508}
}
```
## 参考资料
- [README_ORIGIN](README_ORIGIN.md)
- [https://github.com/facebookresearch/DRTK.git](https://github.com/facebookresearch/DRTK.git)
README_ORIGIN.md 0 → 100644
View file @ d95cd255
# DRTK – Differentiable Rendering Toolkit
This package is a PyTorch library that provides functionality for differentiable rasterization.
It consists of five main components:
* **transform**
* **rasterize**
* **render**
* **interpolate**
* **edge_grad**
There are also optional components such as **msi** and **mipmap_grid_sampler**. New components may be added in the future.
A typical flow looks like this:
**transform** → **rasterize** → **render** → **interpolate** → **CUSTOM SHADING** → **edge_grad**
where:
- **transform**: projects the vertex positions from camera space to image space
- **rasterize**: performs rasterization, where pixels in the output image are associated with triangles
- **render**: computes depth and baricentric image
- **interpolate**: interpolates arbitrary vertex attributes
- **CUSTOM SHADING**: user implemented shading
- **edge_grad**: special module that computes gradients for the **rasterize** step, which is not differentiable on its own. For details, please see [**Rasterized Edge Gradients: Handling Discontinuities Differentiably**](https://arxiv.org/abs/2405.02508)
## Hello Triangle
The "Hello Triangle" with DRTK would look like this:
```python
import drtk
import torch as th
from torchvision.utils import save_image # to save images
# create vertex buffer of shape [1 x n_vertices x 3], here for triangle `n_vertices` == 3
v = th.as_tensor([[[0, 511, 1], [255, 0, 1], [511, 511, 1]]]).float().cuda()
# create index buffer
vi = th.as_tensor([[0, 1, 2]]).int().cuda()
# rasterize
index_img = drtk.rasterize(v, vi, height=512, width=512)
# compute baricentrics
_, bary = drtk.render(v, vi, index_img)
# we won't do shading, we'll just save the baricentrics and filter out the empty region
# which is marked with `-1` in `index_img`
img = bary * (index_img != -1)
save_image(img, "render.png")
```
![hello triangle](docs/source/_static/hellow_triangle.png)
## Dependencies
* PyTorch >= 2.1.0
## Installing
To build a wheel and install it:
```
pip install git+https://github.com/facebookresearch/DRTK.git
```
To build inplace, which is useful for package development:
```
python setup.py build_ext --inplace -j 1
```
## Contributing
See the [CONTRIBUTING](CONTRIBUTING.md) file for how to help out.
## License
DRTK is MIT licensed, as found in the [LICENSE](LICENSE) file.
## Citation
```bibtex
@mish{pidhorskyi2024rasterized,
title = {Rasterized Edge Gradients: Handling Discontinuities Differentiably},
author = {Pidhorskyi, Stanislav and Simon, Tomas and Schwartz, Gabriel and Wen, He and Sheikh, Yaser and Saragih, Jason},
howpublished = {arXiv preprint},
year = {2024},
note = {arXiv:2405.02508}
}
```
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment