p47068915
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

1. edsr network

This commit is contained in:
RiEnRiu 2021-09-03 01:51:41 +08:00
parent 90c7804d6b
commit 58c1588526
2 changed files with 61 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

76
model_zoo/research/cv/EDSR/README_CN.md
View File

@ -38,7 +38,7 @@
增强的深度超分辨率网络(EDSR)是2017年提出的单图超分辨重建网络，在NTIRE2017超分辨重建比赛中获取第一名。它通过删除传统剩余网络中不必要的模块（BatchNorm），扩大模型的大小，同时应用了稳定训练的方法进行优化，显著提升了性能。 增强的深度超分辨率网络(EDSR)是2017年提出的单图超分辨重建网络，在NTIRE2017超分辨重建比赛中获取第一名。它通过删除传统剩余网络中不必要的模块（BatchNorm），扩大模型的大小，同时应用了稳定训练的方法进行优化，显著提升了性能。
[论文](https://arxiv.org/pdf/1707.02921.pdf)：Lim B , Son S , Kim H , et al. Enhanced Deep Residual Networks for Single Image Super-Resolution[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). IEEE, 2017. 论文: [Enhanced Deep Residual Networks for Single Image Super-Resolution](https://arxiv.org/pdf/1707.02921.pdf): Lim B , Son S , Kim H , et al. Enhanced Deep Residual Networks for Single Image Super-Resolution[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). IEEE, 2017.
# 模型架构 # 模型架构
@ -54,6 +54,44 @@ EDSR是由多个优化后的residual blocks串联而成，相比原始版本的r
- 测试集：349.53M，共100组图像(无HR图) - 测试集：349.53M，共100组图像(无HR图)
- 数据格式：PNG图片文件文件 - 数据格式：PNG图片文件文件
- 注：数据将在src/dataset.py中处理。 - 注：数据将在src/dataset.py中处理。
- 数据目录树：官网下载数据后，解压压缩包，训练和验证所需的数据目录结构如下：
```shell
├─DIV2K_train_HR
│ ├─0001.png
│ ├─...
│ └─0800.png
├─DIV2K_train_LR_bicubic
│ ├─X2
│ │ ├─0001x2.png
│ │ ├─...
│ │ └─0800x2.png
│ ├─X3
│ │ ├─0001x3.png
│ │ ├─...
│ │ └─0800x3.png
│ └─X4
│ ├─0001x4.png
│ ├─...
│ └─0800x4.png
├─DIV2K_valid_LR_bicubic
│ ├─0801.png
│ ├─...
│ └─0900.png
└─DIV2K_valid_LR_bicubic
├─X2
│ ├─0801x2.png
│ ├─...
│ └─0900x2.png
├─X3
│ ├─0801x3.png
│ ├─...
│ └─0900x3.png
└─X4
├─0801x4.png
├─...
└─0900x4.png
```
# 特性 # 特性
@ -81,66 +119,66 @@ EDSR是由多个优化后的residual blocks串联而成，相比原始版本的r
```python ```python
# 运行训练示例(EDSR(x2) in the paper) # 运行训练示例(EDSR(x2) in the paper)
python train.py --batch_size 16 --scale 2 --config_path DIV2K_config.yaml > train.log 2>&1 & python train.py --batch_size 16 --config_path DIV2K_config.yaml --scale 2 --data_path [DIV2K path] --output_path [path to save .ckpt] > train.log 2>&1 &
# 运行训练示例(EDSR(x3) in the paper - from EDSR(x2)) # 运行训练示例(EDSR(x3) in the paper - from EDSR(x2))
python train.py --batch_size 16 --scale 3 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x2 model path] train.log 2>&1 & python train.py --batch_size 16 --config_path DIV2K_config.yaml --scale 3 --data_path [DIV2K path] --output_path [path to save .ckpt] --pre_trained [pre-trained EDSR_x2 model path] train.log 2>&1 &
# 运行训练示例(EDSR(x4) in the paper - from EDSR(x2)) # 运行训练示例(EDSR(x4) in the paper - from EDSR(x2))
python train.py --batch_size 16 --scale 4 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x2 model path] train.log 2>&1 & python train.py --batch_size 16 --config_path DIV2K_config.yaml --scale 4 --data_path [DIV2K path] --output_path [path to save .ckpt] --pre_trained [pre-trained EDSR_x2 model path] train.log 2>&1 &
``` ```
- Ascend-910处理器环境运行8卡训练DIV2K - Ascend-910处理器环境运行8卡训练DIV2K
```python ```python
# 运行分布式训练示例(EDSR(x2) in the paper) # 运行分布式训练示例(EDSR(x2) in the paper)
bash scripts/run_train.sh rank_table.json --scale 2 --config_path DIV2K_config.yaml bash scripts/run_train.sh rank_table.json --config_path DIV2K_config.yaml --scale 2 --data_path [DIV2K path] --output_path [path to save .ckpt]
# 运行分布式训练示例(EDSR(x3) in the paper) # 运行分布式训练示例(EDSR(x3) in the paper)
bash scripts/run_train.sh rank_table.json --scale 3 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x2 model path] bash scripts/run_train.sh rank_table.json --config_path DIV2K_config.yaml --scale 3 --data_path [DIV2K path] --output_path [path to save .ckpt] --pre_trained [pre-trained EDSR_x2 model path]
# 运行分布式训练示例(EDSR(x4) in the paper) # 运行分布式训练示例(EDSR(x4) in the paper)
bash scripts/run_train.sh rank_table.json --scale 4 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x2 model path] bash scripts/run_train.sh rank_table.json --config_path DIV2K_config.yaml --scale 4 --data_path [DIV2K path] --output_path [path to save .ckpt] --pre_trained [pre-trained EDSR_x2 model path]
``` ```
- Ascend-910处理器环境运行单卡评估DIV2K - Ascend-910处理器环境运行单卡评估DIV2K
```python ```python
# 运行评估示例(EDSR(x2) in the paper) # 运行评估示例(EDSR(x2) in the paper)
python eval.py --scale 2 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x2 model path] > train.log 2>&1 & python eval.py --config_path DIV2K_config.yaml --scale 2 --data_path [DIV2K path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x2 model path] > train.log 2>&1 &
# 运行评估示例(EDSR(x3) in the paper) # 运行评估示例(EDSR(x3) in the paper)
python eval.py --scale 3 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x3 model path] > train.log 2>&1 & python eval.py --config_path DIV2K_config.yaml --scale 3 --data_path [DIV2K path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x3 model path] > train.log 2>&1 &
# 运行评估示例(EDSR(x4) in the paper) # 运行评估示例(EDSR(x4) in the paper)
python eval.py --scale 4 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x4 model path] > train.log 2>&1 & python eval.py --config_path DIV2K_config.yaml --scale 4 --data_path [DIV2K path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x4 model path] > train.log 2>&1 &
``` ```
- Ascend-910处理器环境运行8卡评估DIV2K - Ascend-910处理器环境运行8卡评估DIV2K
```python ```python
# 运行分布式评估示例(EDSR(x2) in the paper) # 运行分布式评估示例(EDSR(x2) in the paper)
bash scripts/run_eval.sh rank_table.json --scale 2 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x2 model path] bash scripts/run_eval.sh rank_table.json --config_path DIV2K_config.yaml --scale 2 --data_path [DIV2K path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x2 model path]
# 运行分布式评估示例(EDSR(x3) in the paper) # 运行分布式评估示例(EDSR(x3) in the paper)
bash scripts/run_eval.sh rank_table.json --scale 3 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x3 model path] bash scripts/run_eval.sh rank_table.json --config_path DIV2K_config.yaml --scale 3 --data_path [DIV2K path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x3 model path]
# 运行分布式评估示例(EDSR(x4) in the paper) # 运行分布式评估示例(EDSR(x4) in the paper)
bash scripts/run_eval.sh rank_table.json --scale 4 --config_path DIV2K_config.yaml --pre_trained [pre-trained EDSR_x4 model path] bash scripts/run_eval.sh rank_table.json --config_path DIV2K_config.yaml --scale 4 --data_path [DIV2K path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x4 model path]
``` ```
- Ascend-910处理器环境运行单卡评估benchmark - Ascend-910处理器环境运行单卡评估benchmark
```python ```python
# 运行评估示例(EDSR(x2) in the paper) # 运行评估示例(EDSR(x2) in the paper)
python eval.py --scale 2 --config_path benchmark_config.yaml --pre_trained [pre-trained EDSR_x2 model path] > train.log 2>&1 & python eval.py --config_path benchmark_config.yaml --scale 2 --data_path [benchmark path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x2 model path] > train.log 2>&1 &
# 运行评估示例(EDSR(x3) in the paper) # 运行评估示例(EDSR(x3) in the paper)
python eval.py --scale 3 --config_path benchmark_config.yaml --pre_trained [pre-trained EDSR_x3 model path] > train.log 2>&1 & python eval.py --config_path benchmark_config.yaml --scale 3 --data_path [benchmark path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x3 model path] > train.log 2>&1 &
# 运行评估示例(EDSR(x4) in the paper) # 运行评估示例(EDSR(x4) in the paper)
python eval.py --scale 4 --config_path benchmark_config.yaml --pre_trained [pre-trained EDSR_x4 model path] > train.log 2>&1 & python eval.py --config_path benchmark_config.yaml --scale 4 --data_path [benchmark path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x4 model path] > train.log 2>&1 &
``` ```
- Ascend-910处理器环境运行8卡评估benchmark - Ascend-910处理器环境运行8卡评估benchmark
```python ```python
# 运行分布式评估示例(EDSR(x2) in the paper) # 运行分布式评估示例(EDSR(x2) in the paper)
bash scripts/run_eval.sh rank_table.json --scale 2 --config_path benchmark_config.yaml --pre_trained [pre-trained EDSR_x2 model path] bash scripts/run_eval.sh rank_table.json --config_path benchmark_config.yaml --scale 2 --data_path [benchmark path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x2 model path]
# 运行分布式评估示例(EDSR(x3) in the paper) # 运行分布式评估示例(EDSR(x3) in the paper)
bash scripts/run_eval.sh rank_table.json --scale 3 --config_path benchmark_config.yaml --pre_trained [pre-trained EDSR_x3 model path] bash scripts/run_eval.sh rank_table.json --config_path benchmark_config.yaml --scale 3 --data_path [benchmark path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x3 model path]
# 运行分布式评估示例(EDSR(x4) in the paper) # 运行分布式评估示例(EDSR(x4) in the paper)
bash scripts/run_eval.sh rank_table.json --scale 4 --config_path benchmark_config.yaml --pre_trained [pre-trained EDSR_x4 model path] bash scripts/run_eval.sh rank_table.json --config_path benchmark_config.yaml --scale 4 --data_path [benchmark path] --output_path [path to save sr] --pre_trained [pre-trained EDSR_x4 model path]
``` ```
- Ascend-310处理器环境运行单卡评估DIV2K - Ascend-310处理器环境运行单卡评估DIV2K

7
model_zoo/research/cv/EDSR/src/dataset.py
View File

@ -47,8 +47,8 @@ class FolderImagePair:
self.reader = reader self.reader = reader
self.pair_keys, self.image_pairs = self.scan_pair(self.dir_patterns) self.pair_keys, self.image_pairs = self.scan_pair(self.dir_patterns)
@classmethod @staticmethod
def scan_pair(cls, dir_patterns): def scan_pair(dir_patterns):
""" """
scan pair scan pair
""" """
@ -105,7 +105,8 @@ class LrHrImages(FolderImagePair):
def __getitem__(self, idx): def __getitem__(self, idx):
_, *images = super(LrHrImages, self).__getitem__(idx) _, *images = super(LrHrImages, self).__getitem__(idx)
return images return tuple(images)
class _BasePatchCutter: class _BasePatchCutter:
""" """