update Benchmark.md

appendix/benchmark.md

@@ -3,6 +3,7 @@
 ## Photo
 
 Note: waifu2x's photo models was trained on [kou's photo collection](http://photosku.com/photo/category/%E6%92%AE%E5%BD%B1%E8%80%85/kou/).
+
 Note: PSNR in this benchmark uses a [MATLAB's rgb2ycbcr](https://jp.mathworks.com/help/images/ref/rgb2ycbcr.html?lang=en) compatible function (dynamic range [16 235], not [0 255]) for converting grayscale image. I think it's not correct PSNR. But many paper used this metric.
 
 command: 
@@ -10,7 +11,7 @@ command:
 
 ### Datasets
 
-BSD100: https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/segbench/ (100 test images in BSD300)
+BSD100: https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/segbench/ (100 test images in BSDS300)
 Urban100: https://github.com/jbhuang0604/SelfExSR
 
 ### 2x - PSNR