Update README.md
Browse files
README.md
CHANGED
|
@@ -189,6 +189,8 @@ More information needed
|
|
| 189 |
|
| 190 |
The model architecture is based on the original Swin2 arcthitecture for Super Resolution (SR) tasks. The library [transformers](https://github.com/huggingface/transformers) is used to simplify the model design.
|
| 191 |
|
|
|
|
|
|
|
| 192 |
The main component of the model is a [transformers.Swin2SRModel](https://huggingface.co/docs/transformers/model_doc/swin2sr#transformers.Swin2SRModel) which increases x8 the spatial resolution of its inputs (Swin2SR only supports upscaling ratios power of 2).
|
| 193 |
As the real upscale ratio is ~5 and the output shape of the region considered is (160, 240), a Convolutional Neural Network (CNN) is included as a pre-process component which convert the inputs into a (40, 60) feature maps that can be fed to the Swin2SRModel.
|
| 194 |
|
|
|
|
| 189 |
|
| 190 |
The model architecture is based on the original Swin2 arcthitecture for Super Resolution (SR) tasks. The library [transformers](https://github.com/huggingface/transformers) is used to simplify the model design.
|
| 191 |
|
| 192 |
+
<img src="https://huggingface.co/predictia/convswin2sr_mediterranean/blob/main/Diagrama%20%20Swin2SR.png">
|
| 193 |
+
|
| 194 |
The main component of the model is a [transformers.Swin2SRModel](https://huggingface.co/docs/transformers/model_doc/swin2sr#transformers.Swin2SRModel) which increases x8 the spatial resolution of its inputs (Swin2SR only supports upscaling ratios power of 2).
|
| 195 |
As the real upscale ratio is ~5 and the output shape of the region considered is (160, 240), a Convolutional Neural Network (CNN) is included as a pre-process component which convert the inputs into a (40, 60) feature maps that can be fed to the Swin2SRModel.
|
| 196 |
|