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.

Sign up for GitHub

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

Jump to bottom

[Docs] Add Chinese version of customize_models.md #27

Merged
merged 6 commits into from
Feb 26, 2022
Merged

[Docs] Add Chinese version of customize_models.md #27

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
0c43496
[Docs] Add Chinese version of customize_models.md
Abyssaledge Feb 22, 2022
4cdf2bf
[Docs] Add Chinese version of customize_models.md
Abyssaledge Feb 22, 2022
9389ebe
Update customize_models.md
yangxue0827 Feb 23, 2022
6c90aa8
Update customize_models.md
yangxue0827 Feb 23, 2022
d3a0c0b
linting passed
Abyssaledge Feb 23, 2022
1d11853
fixed punctuation
Abyssaledge Feb 24, 2022
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
89 changes: 43 additions & 46 deletions docs/zh_cn/tutorials/customize_models.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,22 +1,22 @@
# Tutorial 3: Customize Models
# 教程 3: 自定义模型

We basically categorize model components into 5 types.
我们大致将模型组件分为了 5 种类型。

- backbone: usually an FCN network to extract feature maps, e.g., ResNet, Swin.
- neck: the component between backbones and heads, e.g., FPN, ReFPN.
- head: the component for specific tasks, e.g., bbox prediction.
- roi extractor: the part for extracting RoI features from feature maps, e.g., RoI Align Rotated.
- loss: the component in head for calculating losses, e.g., FocalLoss, GWDLoss, and KFIoULoss.
- 主干网络 (Backbone): 通常是一个全卷积网络 (FCN)，用来提取特征图，比如残差网络 (ResNet)。也可以是基于视觉 Transformer 的网络，比如 Swin Transformer 等。
- Neck: 主干网络和任务头 (Head) 之间的连接组件，比如 FPN, ReFPN。
- 任务头 (Head): 用于某种具体任务（比如边界框预测）的组件。
- 区域特征提取器 (Roi Extractor): 用于从特征图上提取区域特征的组件，比如 RoI Align Rotated。
- 损失 (loss): 任务头上用于计算损失函数的组件，比如 FocalLoss, GWDLoss, and KFIoULoss。

## Develop new components
## 开发新的组件

### Add a new backbone
### 添加新的主干网络

Here we show how to develop new components with an example of MobileNet.
这里，我们以 MobileNet 为例来展示如何开发新组件。

#### 1. Define a new backbone (e.g. MobileNet)
#### 1. 定义一个新的主干网络（以 MobileNet 为例）

Create a new file `mmrotate/models/backbones/mobilenet.py`.
新建文件 `mmrotate/models/backbones/mobilenet.py`。

```python
import torch.nn as nn
Expand All @@ -34,25 +34,25 @@ class MobileNet(nn.Module):
pass
```

#### 2. Import the module
#### 2. 导入模块

You can either add the following line to `mmrotate/models/backbones/__init__.py`
你可以将下面的代码添加到 `mmrotate/models/backbones/__init__.py` 中：

```python
from .mobilenet import MobileNet
```

or alternatively add
或者添加如下代码

```python
custom_imports = dict(
imports=['mmrotate.models.backbones.mobilenet'],
allow_failed_imports=False)
```

to the config file to avoid modifying the original code.
到配置文件中以避免修改原始代码。

#### 3. Use the backbone in your config file
#### 3. 在你的配置文件中使用该主干网络

```python
model = dict(
Expand All @@ -64,11 +64,11 @@ model = dict(
...
```

### Add new necks
### 添加新的 Neck

#### 1. Define a neck (e.g. PAFPN)
#### 1. 定义一个 Neck（以 PAFPN 为例）

Create a new file `mmrotate/models/necks/pafpn.py`.
新建文件 `mmrotate/models/necks/pafpn.py`。

```python
from mmrotate.models.builder import ROTATED_NECKS
Expand All @@ -90,25 +90,25 @@ class PAFPN(nn.Module):
pass
```

#### 2. Import the module
#### 2. 导入该模块

You can either add the following line to `mmrotate/models/necks/__init__.py`,
你可以添加下述代码到 `mmrotate/models/necks/__init__.py` 中

```python
from .pafpn import PAFPN
```

or alternatively add
或者添加

```python
custom_imports = dict(
imports=['mmrotate.models.necks.pafpn.py'],
allow_failed_imports=False)
```

to the config file and avoid modifying the original code.
到配置文件中以避免修改原始代码。

#### 3. Modify the config file
#### 3. 修改配置文件

```python
neck=dict(
Expand All @@ -118,13 +118,12 @@ neck=dict(
num_outs=5)
```

### Add new heads
### 添加新的 Head

Here we show how to develop a new head with the example of [Double Head R-CNN](https://arxiv.org/abs/1904.06493) as the following.
这里，我们以 [Double Head R-CNN](https://arxiv.org/abs/1904.06493) 为例来展示如何添加一个新的 Head。

First, add a new bbox head in `mmrotate/models/roi_heads/bbox_heads/double_bbox_head.py`.
Double Head R-CNN implements a new bbox head for object detection.
To implement a bbox head, basically we need to implement three functions of the new module as the following.
首先，添加一个新的 bbox head 到 `mmrotate/models/roi_heads/bbox_heads/double_bbox_head.py`。
Double Head R-CNN 在目标检测上实现了一个新的 bbox head。为了实现 bbox head，我们需要使用如下的新模块中三个函数。

```python
from mmrotate.models.builder import ROTATED_HEADS
Expand Down Expand Up @@ -159,7 +158,7 @@ class DoubleConvFCBBoxHead(BBoxHead):

```

Second, implement a new RoI Head if it is necessary. We plan to inherit the new `DoubleHeadRoIHead` from `StandardRoIHead`. We can find that a `StandardRoIHead` already implements the following functions.
然后，如有必要，我们需要实现一个新的 RoI Head。我们打算从 `StandardRoIHead` 继承出新的 `DoubleHeadRoIHead`。我们发现 `StandardRoIHead` 已经实现了下述函数。

```python
import torch
Expand Down Expand Up @@ -206,8 +205,8 @@ class StandardRoIHead(BaseRoIHead, BBoxTestMixin, MaskTestMixin):

```

Double Head's modification is mainly in the bbox_forward logic, and it inherits other logics from the `StandardRoIHead`.
In the `mmrotate/models/roi_heads/double_roi_head.py`, we implement the new RoI Head as the following:
Double Head 的修改主要在 _bbox_forward 的逻辑中，且它从 `StandardRoIHead` 中继承了其他逻辑。
在 `mmrotate/models/roi_heads/double_roi_head.py` 中, 我们实现如下的新的 RoI Head:

```python
from mmrotate.models.builder import ROTATED_HEADS
Expand Down Expand Up @@ -244,24 +243,23 @@ class DoubleHeadRoIHead(StandardRoIHead):
return bbox_results
```

Last, the users need to add the module in
`mmrotate/models/bbox_heads/__init__.py` and `mmrotate/models/roi_heads/__init__.py` thus the corresponding registry could find and load them.
最后，用户需要把这个新模块添加到 `mmrotate/models/bbox_heads/__init__.py` 以及 `mmrotate/models/roi_heads/__init__.py` 中。 这样，注册机制就能找到并加载它们。

Alternatively, the users can add
另外，用户也可以添加

```python
custom_imports=dict(
imports=['mmrotate.models.roi_heads.double_roi_head', 'mmrotate.models.bbox_heads.double_bbox_head'])
```

to the config file and achieve the same goal.
到配置文件中来实现同样的目的。


### Add new loss
### 添加新的损失

Assume you want to add a new loss as `MyLoss`, for bounding box regression.
To add a new loss function, the users need implement it in `mmrotate/models/losses/my_loss.py`.
The decorator `weighted_loss` enable the loss to be weighted for each element.
假设你想添加一个新的损失 `MyLoss` 用于边界框回归。
为了添加一个新的损失函数，用户需要在 `mmrotate/models/losses/my_loss.py` 中实现。
装饰器 `weighted_loss` 可以使损失每个部分加权。

```python
import torch
Expand Down Expand Up @@ -298,24 +296,23 @@ class MyLoss(nn.Module):
return loss_bbox
```

Then the users need to add it in the `mmrotate/models/losses/__init__.py`.
然后，用户需要把下面的代码加到 `mmrotate/models/losses/__init__.py` 中。

```python
from .my_loss import MyLoss, my_loss

```

Alternatively, you can add
或者，你可以添加：

```python
custom_imports=dict(
imports=['mmrotate.models.losses.my_loss'])
```

to the config file and achieve the same goal.
到配置文件来实现相同的目的。

To use it, modify the `loss_xxx` field.
Since MyLoss is for regression, you need to modify the `loss_bbox` field in the head.
因为 MyLoss 是用于回归的，你需要在 Head 中修改 `loss_bbox` 字段：

```python
loss_bbox=dict(type='MyLoss', loss_weight=1.0))
Expand Down