Conversation
|
|
||
| ## 数据准备 | ||
| 模型使用timofte作为训练任务的数据，使用set5或set14作为验证任务的数据。 | ||
| 数据存放结构如下： |
There was a problem hiding this comment.
需要给出数据下载脚本，保证新手可以方便下载到数据，
There was a problem hiding this comment.
需要给出数据下载脚本，保证新手可以方便下载到数据，
已添加，数据集暂无公开的下载链接，所以上传到百度云。
fluid/SRCNN/README.md
Outdated
|
|
||
| 下图是使用`infer.py`脚本预测产生的结果示例，其中，第一行是输入的原始图片，第二行是输入图片经过subsample的结果，第三行是超分辨率重构的结果。 | ||
|
|
||
|  |
There was a problem hiding this comment.
建议将图片放到当前repo下。
已添加到当前repo下
fluid/SRCNN/infer.py
Outdated
|
|
||
| add_arg('checkpoint_path', str, 'model', "Checkpoint save path.") | ||
| add_arg('image_path', str, 'data/val_dataset/set5/baby_GT.bmp', "Img data path.") | ||
| add_arg('show_img', bool, True, "show img or not") |
There was a problem hiding this comment.
默认不要窗口展示图片了，一般用的环境默认没有UI.
There was a problem hiding this comment.
默认不要窗口展示图片了，一般用的环境默认没有UI.
默认设置为False，并修改了这个参数不生效的bug
fluid/SRCNN/train.py
Outdated
|
|
||
| def net(X, Y, model_struct): | ||
| # construct net | ||
| conv1 = fluid.layers.nn.conv2d(X, model_struct.n1, model_struct.f1, act='relu', name='conv1' , |
There was a problem hiding this comment.
fluid.layers.nn.conv2d -> fluid.layers.conv2d
可以不用nn
There was a problem hiding this comment.
fluid.layers.nn.conv2d -> fluid.layers.conv2d
可以不用nn
已修改API
|
稍后我们会将该工作merge到contribute目录。 |
|
|
pancreatic-cystic-lesion
left a comment
pancreatic-cystic-lesion
left a comment
There was a problem hiding this comment.
This is the model of the super-resolution domain, and I'm glad to see the implementation code, but please update the data download link
| ## 数据准备 | ||
| 模型使用timofte作为训练任务的数据，使用set5或set14作为验证任务的数据。数据集获取难度较大，可以通过下面的下载地址下载： | ||
|
|
||
| 链接：https://pan.baidu.com/s/1rynLdqzsbyPamJnKOIwNiA |
There was a problem hiding this comment.
The link has expired, please update
4 participants
add SRCNN code and README.md