Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Docs] Add getting_started chinese version #725

Merged
merged 10 commits into from
Jul 30, 2021
Merged
Show file tree
Hide file tree
Changes from 3 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion docs/getting_started.md
Original file line number Diff line number Diff line change
Expand Up @@ -176,7 +176,7 @@ docker run --gpus all --shm-size=8g -it -v {DATA_DIR}:/mmdetection3d/data mmdete

## A from-scratch setup script

Here is a full script for setting up mmdetection with conda.
Here is a full script for setting up MMdetection3D with conda.

```shell
conda create -n open-mmlab python=3.7 -y
Expand Down
252 changes: 252 additions & 0 deletions docs_zh-CN/getting_started.md
Original file line number Diff line number Diff line change
Expand Up @@ -22,3 +22,255 @@
| 0.5.0 | 2.3.0 | Not required | mmcv-full==1.0.5|

# 安装

xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved
## MMdetection3D 安装流程

**a. 使用 conda 新建虚拟环境,并进入该虚拟环境。**

```shell
conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab
```

**b. 基于 [PyTorch 官网](https://pytorch.org/)安装 PyTorch 和 torchvision,例如:**

```shell
conda install pytorch torchvision -c pytorch
```

**注意**:需要确保 CUDA 的编译版本和运行版本匹配。可以在 [PyTorch 官网](https://pytorch.org/)查看预编译包所支持的 CUDA 版本。

`例 1` 例如在 `/usr/local/cuda` 下安装了 CUDA 10.1, 并想安装 PyTorch 1.5,则需要安装支持 CUDA 10.1 的预构建 PyTorch:

```shell
conda install pytorch cudatoolkit=10.1 torchvision -c pytorch
```

`例 2` 例如在 `/usr/local/cuda` 下安装了 CUDA 9.2, 并想安装 PyTorch 1.3.1,则需要安装支持 CUDA 9.2 的预构建 PyTorch:

```shell
conda install pytorch=1.3.1 cudatoolkit=9.2 torchvision=0.4.2 -c pytorch
```

如果不是安装预构建的包,而是从源码中构建 PyTorch,则可以使用更多的 CUDA 版本,例如 CUDA 9.0。

**c. 安装 [MMCV](https://mmcv.readthedocs.io/en/latest/).**
需要安装 *mmcv-full*, 因为 MMDetection3D 依赖 MMDetection 且需要 *mmcv-full* 中基于 CUDA 的程序.
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

`例` 可以使用下面命令安装预编译版本的 *mmcv-full* : (可使用的版本在[这里](https://mmcv.readthedocs.io/en/latest/#install-with-pip)可以找到)
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

```shell
pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/{cu_version}/{torch_version}/index.html
```
需要把命令行中的 `{cu_version}` 和 `{torch_version}` 替换成对应的版本。例如:在 CUDA 11 和 PyTorch 1.7.0 的环境下,可以使用下面命令安装最新版本的 MMCV:

```shell
pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu110/torch1.7.0/index.html
```

请参考 [MMCV](https://mmcv.readthedocs.io/en/latest/#installation) 获取不同版本的 MMCV 所兼容的的不同的 PyTorch 和 CUDA 版本。同时,也可以通过以下命令行从源码编译 MMCV:

```shell
git clone https://github.com/open-mmlab/mmcv.git
cd mmcv
MMCV_WITH_OPS=1 pip install -e . # 安装好 mmcv-full
cd ..
```

或者,可以直接使用命令行安装:

```shell
pip install mmcv-full
```

**d. 安装 [MMDetection](https://github.com/open-mmlab/mmdetection).**

```shell
pip install mmdet>=2.14.0
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved
```

同时,如果你想修改这部分的代码,也可以通过以下命令从源码编译 MMDetection:

```shell
git clone https://github.com/open-mmlab/mmdetection.git
cd mmdetection
git checkout v2.14.0 # switch to v2.14.0 branch
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved
pip install -r requirements/build.txt
pip install -v -e . # or "python setup.py develop"
```

**e. 安装 [MMSegmentation](https://github.com/open-mmlab/mmsegmentation).**

```shell
pip install mmsegmentation>=0.14.1
ZwwWayne marked this conversation as resolved.
Show resolved Hide resolved
```
同时,如果你想修改这部分的代码,也可以通过以下命令从源码编译 MMSegmentation:

```shell
git clone https://github.com/open-mmlab/mmsegmentation.git
cd mmsegmentation
git checkout v0.14.1 # switch to v0.14.1 branch
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved
pip install -e . # or "python setup.py develop"
```

**g. 安装依赖包和 MMDetection3D.**

```shell
pip install -v -e . # or "python setup.py develop"
```

**注意:**

1. Git 的 commit id 在步骤 d 将会被写入到版本号当中,例 0.6.0+2e7045c 。版本号将保存在训练的模型里。推荐在在每一次执行步骤 d 时,从 github 上获取最新的更新。如果基于 C++/CUDA 的代码被修改了,请执行以下步骤;
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

> 重要: 如果你重装了不同版本的 CUDA 或者 Pythorch 的 mmdet,请务必移除 `./build` 文件。
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

```shell
pip uninstall mmdet3d
rm -rf ./build
find . -name "*.so" | xargs rm
```

2. 按照上述说明,MMDetection 安装在 `dev` 模式下,因此在本地对代码做的任何修改都会生效,无需重新安装;
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

3. 如果希望使用 `opencv-python-headless` 而不是 `opencv-python`, 可以在安装 MMCV 之前安装;

4. 一些安装依赖是可以选择的。例如只需要安装最低运行要求的版本,则可以使用 `pip install -v -e .` 命令。如果希望使用可选择的像 `albumentations` 和 `imagecorruptions` 这种依赖项,可以使用 `pip install -r requirements/optional.txt ` 进行手动安装,或者在使用 `pip` 时指定所需的附加功能(例如 `pip install -v -e .[optional]`),支持附加功能的有效键值包括 `all`、`tests`、`build` 以及 `optional` 。

5. 我们的代码目前不能在只有 CPU 的环境(CUDA 不可用)下编译运行。

## 另一种选择:Docker Image

我们提供了 [Dockerfile](https://github.com/open-mmlab/mmdetection3d/blob/master/docker/Dockerfile) 来建立一个镜像。

```shell
# 基于 PyTorch 1.6, CUDA 10.1 生成 docker 的镜像
docker build -t mmdetection3d docker/
```

运行命令:

```shell
docker run --gpus all --shm-size=8g -it -v {DATA_DIR}:/mmdetection3d/data mmdetection3d
```

## 从零开始的安装脚本

以下是一个基于 conda 安装 MMdetection3D 的脚本

Here is a full script for setting up MMdetection3D with conda.
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

```shell
conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab

# 安装基于环境中默认 CUDA 版本下最新的pytorch (通常使用最新版本)
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved
conda install -c pytorch pytorch torchvision -y

# 安装 mmcv
pip install mmcv-full

# 安装 mmdetection
pip install git+https://github.com/open-mmlab/mmdetection.git

# 安装 mmsegmentation
pip install git+https://github.com/open-mmlab/mmsegmentation.git

# 安装 mmdetection3d
git clone https://github.com/open-mmlab/mmdetection3d.git
cd mmdetection3d
pip install -v -e .
```

## 使用多版本的 MMDetection3D

训练和测试的脚本已经在 PYTHONPATH 中进行了修改,以确保脚本使用当前目录中的 MMDetection3D。

要使环境中安装默认的 MMDetection3D 而不是当前正在在使用的,可以删除出现在相关脚本中的代码:

```shell
PYTHONPATH="$(dirname $0)/..":$PYTHONPATH
```

# 验证

## 通过点云的 demo 来验证
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

我们提供了一些 demo 脚本去测试单个样本,预训练的模型可以从[模型库](model_zoo.md)中下载. 运行如下命令可以去测试点云场景下一个单模态的 3D 检测算法。

```shell
python demo/pcd_demo.py ${PCD_FILE} ${CONFIG_FILE} ${CHECKPOINT_FILE} [--device ${GPU_ID}] [--score-thr ${SCORE_THR}] [--out-dir ${OUT_DIR}]
```

例:

```shell
python demo/pcd_demo.py demo/data/kitti/kitti_000008.bin configs/second/hv_second_secfpn_6x8_80e_kitti-3d-car.py checkpoints/hv_second_secfpn_6x8_80e_kitti-3d-car_20200620_230238-393f000c.pth
```

如果你想输入一个 `ply` 格式的文件,你可以使用如下函数将它转换为 `bin` 的文件格式。然后就可以使用转化成 `bin` 格式的文件去运行 demo。

请注意在使用此脚本前,你需要先安装 pandas 和 plyfile。 这个函数也可使用在数据预处理当中,为了能够直接训练 ```ply data```。
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

```python
import numpy as np
import pandas as pd
from plyfile import PlyData

def convert_ply(input_path, output_path):
plydata = PlyData.read(input_path) # read file
data = plydata.elements[0].data # read data
data_pd = pd.DataFrame(data) # convert to DataFrame
data_np = np.zeros(data_pd.shape, dtype=np.float) # initialize array to store data
property_names = data[0].dtype.names # read names of properties
for i, name in enumerate(
property_names): # read data by property
data_np[:, i] = data_pd[name]
data_np.astype(np.float32).tofile(output_path)
```

例:

```python
convert_ply('./test.ply', './test.bin')
```

如果你有其他格式的点云文件 (例:`off`, `obj`), 你可以使用 trimesh 将它们转化成 `ply`.
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

```python
import trimesh

def to_ply(input_path, output_path, original_type):
mesh = trimesh.load(input_path, file_type=original_type) # read file
mesh.export(output_path, file_type='ply') # convert to ply
```

例:

```python
to_ply('./test.obj', './test.ply', 'obj')
```

更多的关于单/多模态和室内/室外的 3D 检测的样例可以在[此](demo.md)找到.

## 测试点云的高级接口

### 同步接口

这里有一个例子去说明如何构建模型以及测试给出的点云
xiliu8006 marked this conversation as resolved.
Show resolved Hide resolved

```python
from mmdet3d.apis import init_model, inference_detector

config_file = 'configs/votenet/votenet_8x8_scannet-3d-18class.py'
checkpoint_file = 'checkpoints/votenet_8x8_scannet-3d-18class_20200620_230238-2cea9c3a.pth'

# 从配置文件和预训练的模型文件中构建模型
model = init_model(config_file, checkpoint_file, device='cuda:0')

# 测试单个文件并可视化结果
point_cloud = 'test.bin'
result, data = inference_detector(model, point_cloud)
# 可视化结果并且将结果保存到 'results' 文件夹
model.show_results(data, result, out_dir='results')
```