readme

README.md

@@ -20,11 +20,11 @@ git clone https://github.com/geopi1/Improved_USRNet.git
 ### Datasets
 Download the Datasets (a bicubically downscaled version of DIV2K):
 
-[Download DIV2K](https://drive.google.com/drive/folders/1wITaNr7KNg_keUnK4myKjUoKA4HGeONA?usp=sharing) 
-This link will download 2 folders:
-* DIV2K_LR - the low resolution input
-* DIV2K_HR_ds2 - Ground truth high resolution images
+The following links will download the data folders:
+* [DIV2K_HR_ds2](https://technionmail-my.sharepoint.com/:f:/g/personal/pisha_campus_technion_ac_il/EvDI2KscamdIthHRiThIGkIBlDVlADiD6C89g4w0Yr6qbQ?e=Y7MBql) - Ground truth high resolution images
+* [DIV2K_LR](https://technionmail-my.sharepoint.com/:f:/g/personal/pisha_campus_technion_ac_il/EpBY9tozdZhNiM7m1M4yQtsB-GGHnDKHQK_KidTVYpCwVQ?e=u4hhZ1) - the low resolution input
 
+To reproduce the results place the DIV2K_LR images in ./input_images
 ### Prerequisites
 Tested and ran on: 
  * UBUNTU 18.04 
@@ -76,6 +76,24 @@ This will calculate the mean PSNR and SSIM on the SR<->HR on all the results
 | PSNR | 22.35 dB | 22.30 (-0.05) dB | **26.31 (+3.96) dB** | 26.29 (+3.94) dB |
 | SSIM | 0.756 | 0.755 (-0.001) | **0.797 (+0.041)** | 0.797 (+0.041) |
 
+To download the results use the following links:
+
+* [USRGAN_results](https://technionmail-my.sharepoint.com/:f:/g/personal/pisha_campus_technion_ac_il/Evr8_LPl4YREoNp8Lvy4TVcBoXoqrxCB32dQDgDeFz0MQw?e=omnuhY) - USRGAN Results
+* [USRNet_results](https://technionmail-my.sharepoint.com/:f:/g/personal/pisha_campus_technion_ac_il/Ej-ltn6AfO1PlaV1qdXKODABkEGoPZPSnJdQvDB_2OjrWg?e=mSAbKx) - USRNet Results 
+
+Each will download 800 folders where each folder has the following files:
+1. <im_number>_kernel_x2.mat - the estimated kernel
+2. <im_number>\_x4\_<net_type>\_defaultdefualt_noise.png - HR reconstruction with default kernel and default noise levels
+3. <im_number>\_x4\_<net_type>\_defaultdefualt_noise_LE.png - side-by-side of (2) and NN interpolation LR image 
+4. <im_number>\_x4\_<net_type>\_defaultest_noise.png - HR reconstruction with default kernel and estimated noise levels 
+5. <im_number>\_x4\_<net_type>\_defaultdest_noise_LE.png - side-by-side of (4) and NN interpolation LR image 
+6. <im_number>\_x4\_<net_type>\_KernelGANdefault_noise.png - HR reconstruction with estimated kernel and default noise levels 
+7. <im_number>\_x4\_<net_type>\_KernelGANdefault_noise_LE.png - side-by-side of (6) and NN interpolation LR image 
+8. <im_number>\_x4\_<net_type>\_KernelGANest_noise.png - HR reconstruction with estimated kernel and estimated noise levels 
+9. <im_number>\_x4\_<net_type>\_KernelGANest_noise_LE.png - side-by-side of (8) and NN interpolation LR image 
+
+5. psnr_log.txt - psnr & SSIM calculation for each image 
+
 ## Visual Results
 On each image we see (top) the default USRNet estimation with the default kernel, (middle) Blind-USRNet with the estimated kernel and (bottom) Ground Truth. 
 The blur kernel of the LR image is at the top left corner.