Update README.md

README.md

@@ -26,6 +26,18 @@ The dataset contains 19,866 images.
 
 All images are on white background
 
+### Collection Method
+
+Collect from danbooru website.
+
+Use yolov5 to detect and clip image.
+
+Use anime-segmentation to remove background.
+
+Use deepdanbooru to filter image.
+
+Finally clean the dataset manually.
+
 ### Contributions
 
 Thanks to [@SkyTNT](https://github.com/SkyTNT) for adding this dataset.