Eugene Siow commited on
Commit
eea1706
1 Parent(s): cd43d33

Add update to dataset Div2k reference.

Browse files
Files changed (2) hide show
  1. README.md +3 -4
  2. config.json +1 -2
README.md CHANGED
@@ -4,7 +4,7 @@ tags:
4
  - super-image
5
  - image-super-resolution
6
  datasets:
7
- - div2k
8
  metrics:
9
  - pnsr
10
  - ssim
@@ -14,7 +14,7 @@ MSRN model pre-trained on DIV2K (800 images training, augmented to 4000 images,
14
 
15
  The goal of image super resolution is to restore a high resolution (HR) image from a single low resolution (LR) image. The image below shows the ground truth (HR), the bicubic upscaling x2 and model upscaling x2.
16
 
17
- ![Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 4](images/msrn_4_4_compare.png "Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 4")
18
  ## Model description
19
  The MSRN model proposes a feature extraction structure called the multi-scale residual block. This module can "adaptively detect image features at different scales" and "exploit the potential features of the image".
20
 
@@ -83,7 +83,6 @@ training_args = TrainingArguments(
83
  config = MsrnConfig(
84
  scale=4, # train a model to upscale 4x
85
  bam=True, # apply balanced attention to the network
86
- supported_scales=[2, 3, 4],
87
  )
88
  model = MsrnModel(config)
89
 
@@ -122,7 +121,7 @@ The results columns below are represented below as `PSNR/SSIM`. They are compare
122
  |Urban100 |3x | |**29.31/0.8737** |
123
  |Urban100 |4x |23.14/0.6573 |**26.10/0.7857** |
124
 
125
- ![Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 2](images/msrn_2_4_compare.png "Comparing Bicubic upscaling against the models x2 upscaling on Set5 Image 2")
126
 
127
  ## BibTeX entry and citation info
128
  ```bibtex
 
4
  - super-image
5
  - image-super-resolution
6
  datasets:
7
+ - eugenesiow/Div2k
8
  metrics:
9
  - pnsr
10
  - ssim
 
14
 
15
  The goal of image super resolution is to restore a high resolution (HR) image from a single low resolution (LR) image. The image below shows the ground truth (HR), the bicubic upscaling x2 and model upscaling x2.
16
 
17
+ ![Comparing Bicubic upscaling against the models x4 upscaling on Set5 Image 4](images/msrn_4_4_compare.png "Comparing Bicubic upscaling against the models x4 upscaling on Set5 Image 4")
18
  ## Model description
19
  The MSRN model proposes a feature extraction structure called the multi-scale residual block. This module can "adaptively detect image features at different scales" and "exploit the potential features of the image".
20
 
 
83
  config = MsrnConfig(
84
  scale=4, # train a model to upscale 4x
85
  bam=True, # apply balanced attention to the network
 
86
  )
87
  model = MsrnModel(config)
88
 
 
121
  |Urban100 |3x | |**29.31/0.8737** |
122
  |Urban100 |4x |23.14/0.6573 |**26.10/0.7857** |
123
 
124
+ ![Comparing Bicubic upscaling against the models x4 upscaling on Set5 Image 2](images/msrn_2_4_compare.png "Comparing Bicubic upscaling against the models x4 upscaling on Set5 Image 2")
125
 
126
  ## BibTeX entry and citation info
127
  ```bibtex
config.json CHANGED
@@ -5,6 +5,5 @@
5
  "bam": true,
6
  "n_feats": 64,
7
  "n_blocks": 8,
8
- "rgb_range": 255,
9
- "supported_scales": [2,3,4]
10
  }
 
5
  "bam": true,
6
  "n_feats": 64,
7
  "n_blocks": 8,
8
+ "rgb_range": 255
 
9
  }