From 14bafa5e03468775544f8711d7da7a61dbb3d664 Mon Sep 17 00:00:00 2001 From: Xintao Date: Thu, 10 Sep 2020 17:32:44 +0800 Subject: [PATCH 1/7] fixbug: bgr2rgb type conversion in stylegan2 model --- basicsr/models/stylegan2_model.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/basicsr/models/stylegan2_model.py b/basicsr/models/stylegan2_model.py index d3ab279..6bb51ab 100644 --- a/basicsr/models/stylegan2_model.py +++ b/basicsr/models/stylegan2_model.py @@ -1,6 +1,7 @@ import importlib import math import mmcv +import numpy as np import random import torch from collections import OrderedDict @@ -311,7 +312,8 @@ def nondist_validation(self, dataloader, current_iter, tb_logger, f'test_{self.opt["name"]}.png') mmcv.imwrite(result, save_img_path) # add sample images to tb_logger - result = mmcv.bgr2rgb(result / 255.) + result = (result / 255.).astype(np.float32) + result = mmcv.bgr2rgb(result) if tb_logger is not None: tb_logger.add_image( 'samples', result, global_step=current_iter, dataformats='HWC') From ff346f456522907600686655b30637af932ef3d0 Mon Sep 17 00:00:00 2001 From: Xintao Date: Fri, 18 Sep 2020 21:58:44 +0800 Subject: [PATCH 2/7] Add BaiduPan links and use funny emojis (#291) * add links for baidupan * add links for baidupan * update baidupan * test * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README * update README --- README.md | 36 ++++++++++++++++++--------------- README_CN.md | 38 +++++++++++++++++++---------------- docs/Config.md | 2 +- docs/Config_CN.md | 2 +- docs/DatasetPreparation.md | 2 +- docs/DatasetPreparation_CN.md | 2 +- docs/Datasets.md | 2 +- docs/Datasets_CN.md | 2 +- docs/DesignConvention.md | 2 +- docs/DesignConvention_CN.md | 2 +- docs/HOWTOs.md | 8 ++++---- docs/HOWTOs_CN.md | 8 ++++---- docs/Logging.md | 2 +- docs/Logging_CN.md | 2 +- docs/ModelZoo.md | 6 +++--- docs/ModelZoo_CN.md | 7 +++---- docs/Models.md | 2 +- docs/Models_CN.md | 2 +- docs/TrainTest.md | 2 +- docs/TrainTest_CN.md | 2 +- 20 files changed, 69 insertions(+), 62 deletions(-) diff --git a/README.md b/README.md index dd895c4..d17f7bb 100644 --- a/README.md +++ b/README.md @@ -1,7 +1,14 @@ # :rocket: BasicSR -[GitHub](https://github.com/xinntao/BasicSR) | [Gitee码云](https://gitee.com/xinntao/BasicSR)
-[English](README.md) | [简体中文](README_CN.md) +[English](README.md) **|** [简体中文](README_CN.md)   [GitHub](https://github.com/xinntao/BasicSR) **|** [Gitee码云](https://gitee.com/xinntao/BasicSR) + +:arrow_double_down: Google Drive: [Pretrained Models](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) **|** [Reproduced Experiments](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing) +:arrow_double_down: 百度网盘: [预训练模型](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g) **|** [复现实验](https://pan.baidu.com/s/1UElD6q8sVAgn_cxeBDOlvQ)
+:chart_with_upwards_trend: [Training curves in wandb](https://app.wandb.ai/xintao/basicsr)
+:computer: [Commands for training and testing](docs/TrainTest.md)
+:zap: [HOWTOs](#zap-howtos) + +--- BasicSR is an **open source** image and video super-resolution toolbox based on PyTorch (will extend to more restoration tasks in the future).
([ESRGAN](https://github.com/xinntao/ESRGAN), [EDVR](https://github.com/xinntao/EDVR), [DNI](https://github.com/xinntao/DNI), [SFTGAN](https://github.com/xinntao/SFTGAN)) @@ -29,11 +36,11 @@ BasicSR is an **open source** image and video super-resolution toolbox based on We provides simple pipelines to train/test/inference models for quick start. These pipelines/commands cannot cover all the cases and more details are in the following sections. -- :zap: [How to train StyleGAN2](docs/HOWTOs.md#How-to-train-StyleGAN2) -- :zap: [How to test StyleGAN2](docs/HOWTOs.md#How-to-test-StyleGAN2) -- :zap: [How to test DFDNet](docs/HOWTOs.md#How-to-test-DFDNet) +- [How to train StyleGAN2](docs/HOWTOs.md#How-to-train-StyleGAN2) +- [How to test StyleGAN2](docs/HOWTOs.md#How-to-test-StyleGAN2) +- [How to test DFDNet](docs/HOWTOs.md#How-to-test-DFDNet) -## Dependencies and Installation +## :wrench: Dependencies and Installation - Python >= 3.7 (Recommend to use [Anaconda](https://www.anaconda.com/download/#linux) or [Miniconda](https://docs.conda.io/en/latest/miniconda.html)) - [PyTorch >= 1.3](https://pytorch.org/) @@ -49,25 +56,22 @@ python setup.py develop Note that BasicSR is only tested in Ubuntu, and may be not suitable for Windows. You may try [Windows WSL with CUDA supports](https://docs.microsoft.com/en-us/windows/win32/direct3d12/gpu-cuda-in-wsl) :-) (It is now only available for insider build with Fast ring). -## TODO List +## :hourglass_flowing_sand: TODO List Please see [project boards](https://github.com/xinntao/BasicSR/projects). -## Dataset Preparation +## :turtle: Dataset Preparation - Please refer to **[DatasetPreparation.md](docs/DatasetPreparation.md)** for more details. - The descriptions of currently supported datasets (`torch.utils.data.Dataset` classes) are in [Datasets.md](docs/Datasets.md). -## Train and Test +## :computer: Train and Test - **Training and testing commands**: Please see **[TrainTest.md](docs/TrainTest.md)** for the basic usage. - **Options/Configs**: Please refer to [Config.md](docs/Config.md). - **Logging**: Please refer to [Logging.md](docs/Logging.md). -## Model Zoo and Baselines - -**[Download official pre-trained models](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing)**
-**[Download reproduced models and logs](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing)** +## :card_file_box: Model Zoo and Baselines - The descriptions of currently supported models are in [Models.md](docs/Models.md). - **Pre-trained models and log examples** are available in **[ModelZoo.md](docs/ModelZoo.md)**. @@ -78,7 +82,7 @@ Please see [project boards](https://github.com/xinntao/BasicSR/projects).

-## Codebase Designs and Conventions +## :memo: Codebase Designs and Conventions Please see [DesignConvention.md](docs/DesignConvention.md) for the designs and conventions of the BasicSR codebase.
The figure below shows the overall framework. More descriptions for each component:
@@ -86,11 +90,11 @@ The figure below shows the overall framework. More descriptions for each compone ![overall_structure](./assets/overall_structure.png) -## License and Acknowledgement +## :scroll: License and Acknowledgement This project is released under the Apache 2.0 license. More details about license and acknowledgement are in [LICENSE](LICENSE/README.md). -## Contact +## :e-mail: Contact If you have any question, please email `xintao.wang@outlook.com`. diff --git a/README_CN.md b/README_CN.md index 02124d5..f7afe8a 100644 --- a/README_CN.md +++ b/README_CN.md @@ -1,7 +1,14 @@ # :rocket: BasicSR -[GitHub](https://github.com/xinntao/BasicSR) | [Gitee码云](https://gitee.com/xinntao/BasicSR)
-[English](README.md) | [简体中文](README_CN.md) +[English](README.md) **|** [简体中文](README_CN.md)   [GitHub](https://github.com/xinntao/BasicSR) **|** [Gitee码云](https://gitee.com/xinntao/BasicSR) + +:arrow_double_down: 百度网盘: [预训练模型](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g) **|** [复现实验](https://pan.baidu.com/s/1UElD6q8sVAgn_cxeBDOlvQ) +:arrow_double_down: Google Drive: [Pretrained Models](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) **|** [Reproduced Experiments](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing)
+:chart_with_upwards_trend: [wandb的训练曲线](https://app.wandb.ai/xintao/basicsr)
+:computer: [训练和测试的命令](docs/TrainTest_CN.md)
+:zap: [HOWTOs](#zap-howtos) + +--- BasicSR 是一个基于 PyTorch 的**开源**图像视频超分辨率 (Super-Resolution) 工具箱 (之后会支持更多的 Restoration 任务).
([ESRGAN](https://github.com/xinntao/ESRGAN), [EDVR](https://github.com/xinntao/EDVR), [DNI](https://github.com/xinntao/DNI), [SFTGAN](https://github.com/xinntao/SFTGAN)) @@ -24,15 +31,15 @@ BasicSR 是一个基于 PyTorch 的**开源**图像视频超分辨率 (Super-Res -## :zap:HOWTOs +## :zap: HOWTOs 我们提供了简单的流程来快速上手 训练/测试/推理 模型. 这些命令并不能涵盖所有用法, 更多的细节参见下面的部分. -- :zap: [如何训练 StyleGAN2](docs/HOWTOs_CN.md#如何训练-StyleGAN2) -- :zap: [如何测试 StyleGAN2](docs/HOWTOs_CN.md#如何测试-StyleGAN2) -- :zap: [如何测试 DFDNet](docs/HOWTOs_CN.md#如何测试-DFDNet) +- [如何训练 StyleGAN2](docs/HOWTOs_CN.md#如何训练-StyleGAN2) +- [如何测试 StyleGAN2](docs/HOWTOs_CN.md#如何测试-StyleGAN2) +- [如何测试 DFDNet](docs/HOWTOs_CN.md#如何测试-DFDNet) -## 依赖和安装 +## :wrench: 依赖和安装 - Python >= 3.7 (推荐使用 [Anaconda](https://www.anaconda.com/download/#linux) 或 [Miniconda](https://docs.conda.io/en/latest/miniconda.html)) - [PyTorch >= 1.3](https://pytorch.org/) @@ -48,25 +55,22 @@ python setup.py develop 注意: BasicSR 仅在 Ubuntu 下进行测试，或许不支持Windows. 可以在Windows下尝试[支持CUDA的Windows WSL](https://docs.microsoft.com/en-us/windows/win32/direct3d12/gpu-cuda-in-wsl) :-) (目前只有Fast ring的预览版系统可以安装). -## TODO 清单 +## :hourglass_flowing_sand: TODO 清单 参见 [project boards](https://github.com/xinntao/BasicSR/projects). -## 数据准备 +## :turtle: 数据准备 - 数据准备步骤, 参见 **[DatasetPreparation_CN.md](docs/DatasetPreparation_CN.md)**. - 目前支持的数据集 (`torch.utils.data.Dataset`类), 参见 [Datasets_CN.md](docs/Datasets_CN.md). -## 训练和测试 +## :computer: 训练和测试 - **训练和测试的命令**, 参见 **[TrainTest_CN.md](docs/TrainTest_CN.md)**. - **Options/Configs**配置文件的说明, 参见 [Config_CN.md](docs/Config_CN.md). - **Logging**日志系统的说明, 参见 [Logging_CN.md](docs/Logging_CN.md). -## 模型库和基准 - -**[下载官方提供的预训练模型](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing)**
-**[下载复现的模型和log](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing)** +## :card_file_box: 模型库和基准 - 目前支持的模型描述, 参见 [Models_CN.md](docs/Models_CN.md). - **预训练模型和log样例**, 参见 **[ModelZoo_CN.md](docs/ModelZoo_CN.md)**. @@ -77,7 +81,7 @@ python setup.py develop

-## 代码库的设计和约定 +## :memo: 代码库的设计和约定 参见 [DesignConvention_CN.md](docs/DesignConvention_CN.md).
下图概括了整体的框架. 每个模块更多的描述参见:
@@ -85,11 +89,11 @@ python setup.py develop ![overall_structure](./assets/overall_structure.png) -## 许可 +## :scroll: 许可 本项目使用 Apache 2.0 license. 更多细节参见 [LICENSE](LICENSE/README.md). -#### 联系 +## :e-mail: 联系 若有任何问题, 请电邮 `xintao.wang@outlook.com`. diff --git a/docs/Config.md b/docs/Config.md index 3c0163b..f2a0775 100644 --- a/docs/Config.md +++ b/docs/Config.md @@ -1,6 +1,6 @@ # Configuration -[English](Config.md) | [简体中文](Config_CN.md) +[English](Config.md) **|** [简体中文](Config_CN.md) #### Contents diff --git a/docs/Config_CN.md b/docs/Config_CN.md index db77fc6..6fa159d 100644 --- a/docs/Config_CN.md +++ b/docs/Config_CN.md @@ -1,6 +1,6 @@ # 配置文件 -[English](Config.md) | [简体中文](Config_CN.md) +[English](Config.md) **|** [简体中文](Config_CN.md) #### 目录 diff --git a/docs/DatasetPreparation.md b/docs/DatasetPreparation.md index a18159d..b579f58 100644 --- a/docs/DatasetPreparation.md +++ b/docs/DatasetPreparation.md @@ -1,6 +1,6 @@ # Dataset Preparation -[English](DatasetPreparation.md) | [简体中文](DatasetPreparation_CN.md) +[English](DatasetPreparation.md) **|** [简体中文](DatasetPreparation_CN.md) #### Contents diff --git a/docs/DatasetPreparation_CN.md b/docs/DatasetPreparation_CN.md index b8dc492..7600256 100644 --- a/docs/DatasetPreparation_CN.md +++ b/docs/DatasetPreparation_CN.md @@ -1,6 +1,6 @@ # 数据准备 -[English](DatasetPreparation.md) | [简体中文](DatasetPreparation_CN.md) +[English](DatasetPreparation.md) **|** [简体中文](DatasetPreparation_CN.md) #### 目录 diff --git a/docs/Datasets.md b/docs/Datasets.md index 2264823..c7653af 100644 --- a/docs/Datasets.md +++ b/docs/Datasets.md @@ -1,6 +1,6 @@ # Datasets -[English](Datasets.md) | [简体中文](Datasets_CN.md) +[English](Datasets.md) **|** [简体中文](Datasets_CN.md) ## Supported Datasets diff --git a/docs/Datasets_CN.md b/docs/Datasets_CN.md index e785f2c..0a6324a 100644 --- a/docs/Datasets_CN.md +++ b/docs/Datasets_CN.md @@ -1,6 +1,6 @@ # 数据处理 -[English](Datasets.md) | [简体中文](Datasets_CN.md) +[English](Datasets.md) **|** [简体中文](Datasets_CN.md) ## 支持的数据处理 diff --git a/docs/DesignConvention.md b/docs/DesignConvention.md index 4828147..35ee55f 100644 --- a/docs/DesignConvention.md +++ b/docs/DesignConvention.md @@ -1,6 +1,6 @@ # Codebase Designs and Conventions -[English](DesignConvention.md) | [简体中文](DesignConvention_CN.md) +[English](DesignConvention.md) **|** [简体中文](DesignConvention_CN.md) #### Contents diff --git a/docs/DesignConvention_CN.md b/docs/DesignConvention_CN.md index 8b20111..d3c16d3 100644 --- a/docs/DesignConvention_CN.md +++ b/docs/DesignConvention_CN.md @@ -1,6 +1,6 @@ # 代码库的设计和约定 -[English](DesignConvention.md) | [简体中文](DesignConvention_CN.md) +[English](DesignConvention.md) **|** [简体中文](DesignConvention_CN.md) #### 目录 diff --git a/docs/HOWTOs.md b/docs/HOWTOs.md index 1b2632c..b698872 100644 --- a/docs/HOWTOs.md +++ b/docs/HOWTOs.md @@ -1,6 +1,6 @@ # HOWTOs -[English](HOWTOs.md) | [简体中文](HOWTOs_CN.md) +[English](HOWTOs.md) **|** [简体中文](HOWTOs_CN.md) ## How to train StyleGAN2 @@ -17,7 +17,7 @@ ## How to test StyleGAN2 -1. Download pre-trained models from [ModelZoo](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) to the `experiments/pretrained_models` folder. +1. Download pre-trained models from **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) to the `experiments/pretrained_models` folder. 1. Test. > python tests/test_stylegan2.py @@ -30,12 +30,12 @@ 1. Clone dlib repo: `git clone git@github.com:davisking/dlib.git` 1. `cd dlib` 1. Install: `python setup.py install` -2. Download the dlib pretrained models from [ModelZoo](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) to the `experiments/pretrained_models/dlib` folder.
+2. Download the dlib pretrained models from **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) to the `experiments/pretrained_models/dlib` folder.
You can download by run the following command OR manually download the pretrained models. > python scripts/download_pretrained_models.py --method dlib -3. Download pretrained DFDNet models, dictionary and face template from [ModelZoo](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) to the `experiments/pretrained_models/DFDNet` folder.
+3. Download pretrained DFDNet models, dictionary and face template from **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) to the `experiments/pretrained_models/DFDNet` folder.
You can download by run the the following command OR manually download the pretrained models. > python scripts/download_pretrained_models.py --method DFDNet diff --git a/docs/HOWTOs_CN.md b/docs/HOWTOs_CN.md index e728e53..3ef2921 100644 --- a/docs/HOWTOs_CN.md +++ b/docs/HOWTOs_CN.md @@ -1,6 +1,6 @@ # HOWTOs -[English](HOWTOs.md) | [简体中文](HOWTOs_CN.md) +[English](HOWTOs.md) **|** [简体中文](HOWTOs_CN.md) ## 如何训练 StyleGAN2 @@ -17,7 +17,7 @@ ## 如何测试 StyleGAN2 -1. 从 [ModelZoo](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) 下载预训练模型到 `experiments/pretrained_models` 文件夹. +1. 从 **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) 下载预训练模型到 `experiments/pretrained_models` 文件夹. 1. 测试. > python tests/test_stylegan2.py @@ -30,12 +30,12 @@ 1. 克隆 dlib repo: `git clone git@github.com:davisking/dlib.git` 1. `cd dlib` 1. 安装: `python setup.py install` -2. 从 [ModelZoo](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) 下载预训练的 dlib 模型到 `experiments/pretrained_models/dlib` 文件夹.
+2. 从 **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) 下载预训练的 dlib 模型到 `experiments/pretrained_models/dlib` 文件夹.
你可以通过运行下面的命令下载 或 手动下载. > python scripts/download_pretrained_models.py --method dlib -3. 从 [ModelZoo](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing) 下载 DFDNet 模型, 字典和人脸关键点模板到 `experiments/pretrained_models/DFDNet` 文件夹.
+3. 从 **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) 下载 DFDNet 模型, 字典和人脸关键点模板到 `experiments/pretrained_models/DFDNet` 文件夹.
你可以通过运行下面的命令下载 或 手动下载. > python scripts/download_pretrained_models.py --method DFDNet diff --git a/docs/Logging.md b/docs/Logging.md index 68e486b..ac2e45c 100644 --- a/docs/Logging.md +++ b/docs/Logging.md @@ -1,6 +1,6 @@ # Logging -[English](Logging.md) | [简体中文](Logging_CN.md) +[English](Logging.md) **|** [简体中文](Logging_CN.md) #### Contents diff --git a/docs/Logging_CN.md b/docs/Logging_CN.md index 01d489e..cb1889f 100644 --- a/docs/Logging_CN.md +++ b/docs/Logging_CN.md @@ -1,6 +1,6 @@ # Logging日志 -[English](Logging.md) | [简体中文](Logging_CN.md) +[English](Logging.md) **|** [简体中文](Logging_CN.md) #### 目录 diff --git a/docs/ModelZoo.md b/docs/ModelZoo.md index 4431ec5..aa012d0 100644 --- a/docs/ModelZoo.md +++ b/docs/ModelZoo.md @@ -1,6 +1,6 @@ # Model Zoo and Baselines -[English](ModelZoo.md) | [简体中文](ModelZoo_CN.md) +[English](ModelZoo.md) **|** [简体中文](ModelZoo_CN.md) We provide: @@ -9,7 +9,7 @@ We provide: You can put the downloaded models in the `experiments/pretrained_models` folder. -**[Download official pre-trained models](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing)** +**[Download official pre-trained models]** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g))(https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing)) You can use the scrip to download pre-trained models from Google Drive. @@ -18,7 +18,7 @@ python scripts/download_pretrained_models.py --method ESRGAN # method can be ESRGAN, EDVR, StyleGAN, EDSR, DUF, DFDNet, dlib ``` -**[Download reproduced models and logs](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing)** +**[Download reproduced models and logs]** ([Google Drive](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing), [百度网盘](https://pan.baidu.com/s/1UElD6q8sVAgn_cxeBDOlvQ)) In addition, we upload the training process and curves in [wandb](https://www.wandb.com/). diff --git a/docs/ModelZoo_CN.md b/docs/ModelZoo_CN.md index 8299231..bcb4fbf 100644 --- a/docs/ModelZoo_CN.md +++ b/docs/ModelZoo_CN.md @@ -1,6 +1,6 @@ # 模型库和基准 -[English](ModelZoo.md) | [简体中文](ModelZoo_CN.md) +[English](ModelZoo.md) **|** [简体中文](ModelZoo_CN.md) 我们提供了: @@ -9,8 +9,7 @@ 下载的模型可以放在 `experiments/pretrained_models` 文件夹. -**[下载官方提供的预训练模型](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing)** - +**[下载官方提供的预训练模型]** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) 你可以使用以下脚本从Google Drive下载预训练模型. ```python @@ -18,7 +17,7 @@ python scripts/download_pretrained_models.py --method ESRGAN # method can be ESRGAN, EDVR, StyleGAN, EDSR, DUF, DFDNet, dlib ``` -**[下载复现的模型和log](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing)** +**[下载复现的模型和log]** ([Google Drive](https://drive.google.com/drive/folders/1XN4WXKJ53KQ0Cu0Yv-uCt8DZWq6uufaP?usp=sharing), [百度网盘](https://pan.baidu.com/s/1UElD6q8sVAgn_cxeBDOlvQ)) 此外, 我们在 [wandb](https://www.wandb.com/) 上更新了模型训练的过程和曲线. 大家可以方便的比较: diff --git a/docs/Models.md b/docs/Models.md index bf6e96d..35e156a 100644 --- a/docs/Models.md +++ b/docs/Models.md @@ -1,6 +1,6 @@ # Models -[English](Models.md) | [简体中文](Models_CN.md) +[English](Models.md) **|** [简体中文](Models_CN.md) #### Contents diff --git a/docs/Models_CN.md b/docs/Models_CN.md index 7f95376..35eee42 100644 --- a/docs/Models_CN.md +++ b/docs/Models_CN.md @@ -1,6 +1,6 @@ # 模型 -[English](Models.md) | [简体中文](Models_CN.md) +[English](Models.md) **|** [简体中文](Models_CN.md) #### 目录 diff --git a/docs/TrainTest.md b/docs/TrainTest.md index 4a6cab2..b3d7cd7 100644 --- a/docs/TrainTest.md +++ b/docs/TrainTest.md @@ -1,6 +1,6 @@ # Training and Testing -[English](TrainTest.md) | [简体中文](TrainTest_CN.md) +[English](TrainTest.md) **|** [简体中文](TrainTest_CN.md) Please run the commands in the root path of `BasicSR`.
In general, both the training and testing include the following steps: diff --git a/docs/TrainTest_CN.md b/docs/TrainTest_CN.md index 2bb66e0..21c4cbd 100644 --- a/docs/TrainTest_CN.md +++ b/docs/TrainTest_CN.md @@ -1,6 +1,6 @@ # 训练和测试 -[English](TrainTest.md) | [简体中文](TrainTest_CN.md) +[English](TrainTest.md) **|** [简体中文](TrainTest_CN.md) 所有的命令都在 `BasicSR` 的根目录下运行.
一般来说, 训练和测试都有以下的步骤: From 56021e43d935c9b38f0f7d0f93e9f4c4a928a4c7 Mon Sep 17 00:00:00 2001 From: Xintao Date: Fri, 18 Sep 2020 23:44:30 +0800 Subject: [PATCH 3/7] download_pretrained_models support download all the models --- docs/HOWTOs.md | 4 ++-- docs/HOWTOs_CN.md | 4 ++-- docs/ModelZoo.md | 2 +- docs/ModelZoo_CN.md | 2 +- scripts/download_pretrained_models.py | 13 +++++++++++-- 5 files changed, 17 insertions(+), 8 deletions(-) diff --git a/docs/HOWTOs.md b/docs/HOWTOs.md index b698872..a2d6433 100644 --- a/docs/HOWTOs.md +++ b/docs/HOWTOs.md @@ -33,12 +33,12 @@ 2. Download the dlib pretrained models from **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) to the `experiments/pretrained_models/dlib` folder.
You can download by run the following command OR manually download the pretrained models. - > python scripts/download_pretrained_models.py --method dlib + > python scripts/download_pretrained_models.py dlib 3. Download pretrained DFDNet models, dictionary and face template from **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) to the `experiments/pretrained_models/DFDNet` folder.
You can download by run the the following command OR manually download the pretrained models. - > python scripts/download_pretrained_models.py --method DFDNet + > python scripts/download_pretrained_models.py DFDNet 4. Prepare the testing dataset in the `datasets`, for example, we put images in the `datasets/TestWhole` folder. 5. Test. diff --git a/docs/HOWTOs_CN.md b/docs/HOWTOs_CN.md index 3ef2921..aad7f25 100644 --- a/docs/HOWTOs_CN.md +++ b/docs/HOWTOs_CN.md @@ -33,12 +33,12 @@ 2. 从 **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) 下载预训练的 dlib 模型到 `experiments/pretrained_models/dlib` 文件夹.
你可以通过运行下面的命令下载 或 手动下载. - > python scripts/download_pretrained_models.py --method dlib + > python scripts/download_pretrained_models.py dlib 3. 从 **ModelZoo** ([Google Drive](https://drive.google.com/drive/folders/15DgDtfaLASQ3iAPJEVHQF49g9msexECG?usp=sharing), [百度网盘](https://pan.baidu.com/s/1R6Nc4v3cl79XPAiK0Toe7g)) 下载 DFDNet 模型, 字典和人脸关键点模板到 `experiments/pretrained_models/DFDNet` 文件夹.
你可以通过运行下面的命令下载 或 手动下载. - > python scripts/download_pretrained_models.py --method DFDNet + > python scripts/download_pretrained_models.py DFDNet 4. 准备测试图片到 `datasets`, 比如说我们把测试图片放在 `datasets/TestWhole` 文件夹. 5. 测试. diff --git a/docs/ModelZoo.md b/docs/ModelZoo.md index aa012d0..af6579c 100644 --- a/docs/ModelZoo.md +++ b/docs/ModelZoo.md @@ -14,7 +14,7 @@ You can put the downloaded models in the `experiments/pretrained_models` folder. You can use the scrip to download pre-trained models from Google Drive. ```python -python scripts/download_pretrained_models.py --method ESRGAN +python scripts/download_pretrained_models.py ESRGAN # method can be ESRGAN, EDVR, StyleGAN, EDSR, DUF, DFDNet, dlib ``` diff --git a/docs/ModelZoo_CN.md b/docs/ModelZoo_CN.md index bcb4fbf..9290a91 100644 --- a/docs/ModelZoo_CN.md +++ b/docs/ModelZoo_CN.md @@ -13,7 +13,7 @@ 你可以使用以下脚本从Google Drive下载预训练模型. ```python -python scripts/download_pretrained_models.py --method ESRGAN +python scripts/download_pretrained_models.py ESRGAN # method can be ESRGAN, EDVR, StyleGAN, EDSR, DUF, DFDNet, dlib ``` diff --git a/scripts/download_pretrained_models.py b/scripts/download_pretrained_models.py index a2fec5d..cc26218 100644 --- a/scripts/download_pretrained_models.py +++ b/scripts/download_pretrained_models.py @@ -29,7 +29,12 @@ def download_pretrained_models(method, file_ids): if __name__ == '__main__': parser = argparse.ArgumentParser() - parser.add_argument('--method', type=str, default='ESRGAN') + parser.add_argument( + 'method', + type=str, + help=( + "Options: 'ESRGAN', 'EDVR', 'StyleGAN', 'EDSR', 'DUF', 'DFDNet', " + "'dlib'. Set to 'all' if you want to download all the models.")) args = parser.parse_args() file_ids = { @@ -121,4 +126,8 @@ def download_pretrained_models(method, file_ids): } } - download_pretrained_models(args.method, file_ids[args.method]) + if args.method == 'all': + for method in file_ids.keys(): + download_pretrained_models(method, file_ids[method]) + else: + download_pretrained_models(args.method, file_ids[args.method]) From 3ea61c637abf37b269ee50a55ebe42a8ae557563 Mon Sep 17 00:00:00 2001 From: Xintao Date: Fri, 18 Sep 2020 23:45:43 +0800 Subject: [PATCH 4/7] support training w/o val --- basicsr/train.py | 6 +++--- basicsr/utils/options.py | 3 ++- 2 files changed, 5 insertions(+), 4 deletions(-) diff --git a/basicsr/train.py b/basicsr/train.py index 07c0736..0d769c8 100644 --- a/basicsr/train.py +++ b/basicsr/train.py @@ -205,8 +205,8 @@ def main(): model.save(epoch, current_iter) # validation - if opt['val']['val_freq'] is not None and current_iter % opt[ - 'val']['val_freq'] == 0: + if opt.get('val') is not None and (current_iter % + opt['val']['val_freq'] == 0): model.validation(val_loader, current_iter, tb_logger, opt['val']['save_img']) @@ -222,7 +222,7 @@ def main(): logger.info(f'End of training. Time consumed: {consumed_time}') logger.info('Save the latest model.') model.save(epoch=-1, current_iter=-1) # -1 stands for the latest - if opt['val']['val_freq'] is not None: + if opt.get('val') is not None: model.validation(val_loader, current_iter, tb_logger, opt['val']['save_img']) if tb_logger: diff --git a/basicsr/utils/options.py b/basicsr/utils/options.py index cddf25e..f7717f0 100644 --- a/basicsr/utils/options.py +++ b/basicsr/utils/options.py @@ -75,7 +75,8 @@ def parse(opt_path, is_train=True): # change some options for debug mode if 'debug' in opt['name']: - opt['val']['val_freq'] = 8 + if 'val' in opt: + opt['val']['val_freq'] = 8 opt['logger']['print_freq'] = 1 opt['logger']['save_checkpoint_freq'] = 8 else: # test From f1c1febc37dbc6697822c65ca53d6dffb6f47015 Mon Sep 17 00:00:00 2001 From: Xintao Date: Fri, 18 Sep 2020 23:47:46 +0800 Subject: [PATCH 5/7] reorganize define_network --- basicsr/models/archs/__init__.py | 28 ++++++++++++++++++++++ basicsr/models/networks.py | 39 ------------------------------- basicsr/models/sr_model.py | 4 ++-- basicsr/models/srgan_model.py | 4 ++-- basicsr/models/stylegan2_model.py | 13 ++++++----- basicsr/models/video_gan_model.py | 4 ++-- 6 files changed, 41 insertions(+), 51 deletions(-) delete mode 100644 basicsr/models/networks.py diff --git a/basicsr/models/archs/__init__.py b/basicsr/models/archs/__init__.py index 8f17af5..a00982a 100644 --- a/basicsr/models/archs/__init__.py +++ b/basicsr/models/archs/__init__.py @@ -15,3 +15,31 @@ importlib.import_module(f'basicsr.models.archs.{file_name}') for file_name in arch_filenames ] + + +def dynamic_instantiation(modules, cls_type, opt): + """Dynamically instantiate class. + + Args: + modules (list[importlib modules]): List of modules from importlib + files. + cls_type (str): Class type. + opt (dict): Class initialization kwargs. + + Returns: + class: Instantiated class. + """ + + for module in modules: + cls_ = getattr(module, cls_type, None) + if cls_ is not None: + break + if cls_ is None: + raise ValueError(f'{cls_type} is not found.') + return cls_(**opt) + + +def define_network(opt): + network_type = opt.pop('type') + net = dynamic_instantiation(_arch_modules, network_type, opt) + return net diff --git a/basicsr/models/networks.py b/basicsr/models/networks.py deleted file mode 100644 index ecc97f1..0000000 --- a/basicsr/models/networks.py +++ /dev/null @@ -1,39 +0,0 @@ -from basicsr.models.archs import _arch_modules - - -def dynamic_instantiation(modules, cls_type, opt): - """Dynamically instantiate class. - - Args: - modules (list[importlib modules]): List of modules from importlib - files. - cls_type (str): Class type. - opt (dict): Class initialization kwargs. - - Returns: - class： Instantiated class. - """ - - for module in modules: - cls_ = getattr(module, cls_type, None) - if cls_ is not None: - break - if cls_ is None: - raise ValueError(f'{cls_type} is not found.') - return cls_(**opt) - - -# Generator -def define_net_g(opt): - network_type = opt.pop('type') - net_g = dynamic_instantiation(_arch_modules, network_type, opt) - - return net_g - - -# Discriminator -def define_net_d(opt): - network_type = opt.pop('type') - - net_d = dynamic_instantiation(_arch_modules, network_type, opt) - return net_d diff --git a/basicsr/models/sr_model.py b/basicsr/models/sr_model.py index 56b6193..66a98b6 100644 --- a/basicsr/models/sr_model.py +++ b/basicsr/models/sr_model.py @@ -5,7 +5,7 @@ from copy import deepcopy from os import path as osp -from basicsr.models import networks as networks +from basicsr.models.archs import define_network from basicsr.models.base_model import BaseModel from basicsr.utils import ProgressBar, get_root_logger, tensor2img @@ -20,7 +20,7 @@ def __init__(self, opt): super(SRModel, self).__init__(opt) # define network - self.net_g = networks.define_net_g(deepcopy(opt['network_g'])) + self.net_g = define_network(deepcopy(opt['network_g'])) self.net_g = self.model_to_device(self.net_g) self.print_network(self.net_g) diff --git a/basicsr/models/srgan_model.py b/basicsr/models/srgan_model.py index a9e7c27..d927773 100644 --- a/basicsr/models/srgan_model.py +++ b/basicsr/models/srgan_model.py @@ -3,7 +3,7 @@ from collections import OrderedDict from copy import deepcopy -from basicsr.models import networks as networks +from basicsr.models.archs import define_network from basicsr.models.sr_model import SRModel loss_module = importlib.import_module('basicsr.models.losses') @@ -16,7 +16,7 @@ def init_training_settings(self): train_opt = self.opt['train'] # define network net_d - self.net_d = networks.define_net_d(deepcopy(self.opt['network_d'])) + self.net_d = define_network(deepcopy(self.opt['network_d'])) self.net_d = self.model_to_device(self.net_d) self.print_network(self.net_d) diff --git a/basicsr/models/stylegan2_model.py b/basicsr/models/stylegan2_model.py index 6bb51ab..7cf7aec 100644 --- a/basicsr/models/stylegan2_model.py +++ b/basicsr/models/stylegan2_model.py @@ -8,7 +8,7 @@ from copy import deepcopy from os import path as osp -from basicsr.models import networks as networks +from basicsr.models.archs import define_network from basicsr.models.base_model import BaseModel from basicsr.models.losses.losses import g_path_regularize, r1_penalty from basicsr.utils import tensor2img @@ -23,7 +23,7 @@ def __init__(self, opt): super(StyleGAN2Model, self).__init__(opt) # define network net_g - self.net_g = networks.define_net_g(deepcopy(opt['network_g'])) + self.net_g = define_network(deepcopy(opt['network_g'])) self.net_g = self.model_to_device(self.net_g) self.print_network(self.net_g) # load pretrained model @@ -35,8 +35,9 @@ def __init__(self, opt): # latent dimension: self.num_style_feat self.num_style_feat = opt['network_g']['num_style_feat'] + num_val_samples = self.opt['val'].get('num_val_samples', 16) self.fixed_sample = torch.randn( - 16, self.num_style_feat, device=self.device) + num_val_samples, self.num_style_feat, device=self.device) if self.is_train: self.init_training_settings() @@ -45,7 +46,7 @@ def init_training_settings(self): train_opt = self.opt['train'] # define network net_d - self.net_d = networks.define_net_d(deepcopy(self.opt['network_d'])) + self.net_d = define_network(deepcopy(self.opt['network_d'])) self.net_d = self.model_to_device(self.net_d) self.print_network(self.net_d) @@ -58,8 +59,8 @@ def init_training_settings(self): # define network net_g with Exponential Moving Average (EMA) # net_g_ema only used for testing on one GPU and saving, do not need to # wrap with DistributedDataParallel - self.net_g_ema = networks.define_net_g( - deepcopy(self.opt['network_g'])).to(self.device) + self.net_g_ema = define_network(deepcopy(self.opt['network_g'])).to( + self.device) # load pretrained model load_path = self.opt['path'].get('pretrain_model_g', None) if load_path is not None: diff --git a/basicsr/models/video_gan_model.py b/basicsr/models/video_gan_model.py index 836954b..94ccf4b 100644 --- a/basicsr/models/video_gan_model.py +++ b/basicsr/models/video_gan_model.py @@ -3,7 +3,7 @@ from collections import OrderedDict from copy import deepcopy -from basicsr.models import networks as networks +from basicsr.models.archs import define_network from basicsr.models.video_base_model import VideoBaseModel loss_module = importlib.import_module('basicsr.models.losses') @@ -16,7 +16,7 @@ def init_training_settings(self): train_opt = self.opt['train'] # define network net_d - self.net_d = networks.define_net_d(deepcopy(self.opt['network_d'])) + self.net_d = define_network(deepcopy(self.opt['network_d'])) self.net_d = self.model_to_device(self.net_d) self.print_network(self.net_d) From 18c1da087e270026da7ee9d8c697fed1172cf98c Mon Sep 17 00:00:00 2001 From: Xintao Date: Fri, 18 Sep 2020 23:48:26 +0800 Subject: [PATCH 6/7] add paths_from_folder to data util --- basicsr/data/util.py | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/basicsr/data/util.py b/basicsr/data/util.py index 798d368..50245ac 100644 --- a/basicsr/data/util.py +++ b/basicsr/data/util.py @@ -246,6 +246,21 @@ def paired_paths_from_folder(folders, keys, filename_tmpl): return paths +def paths_from_folder(folder): + """Generate paths from folder. + + Args: + folder (str): Folder path. + + Returns: + list[str]: Returned path list. + """ + + paths = list(mmcv.scandir(folder)) + paths = [osp.join(folder, path) for path in paths] + return paths + + def generate_gaussian_kernel(kernel_size=13, sigma=1.6): """Generate Gaussian kernel used in `duf_downsample`. From b25509ec29a38ae8ed8f8d5e5873d3e719d2a33a Mon Sep 17 00:00:00 2001 From: Xintao Date: Fri, 18 Sep 2020 23:50:46 +0800 Subject: [PATCH 7/7] :bookmark: VERSION 1.1.1 --- VERSION | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/VERSION b/VERSION index 9084fa2..524cb55 100644 --- a/VERSION +++ b/VERSION @@ -1 +1 @@ -1.1.0 +1.1.1