create README

README.md

@@ -5,13 +5,13 @@ dataset_info:
     dtype: image
   splits:
   - name: train
-    num_bytes: 24475559.0
+    num_bytes: 24475559
     num_examples: 337
   - name: val
-    num_bytes: 11307736.0
+    num_bytes: 11307736
     num_examples: 113
   download_size: 35763051
-  dataset_size: 35783295.0
+  dataset_size: 35783295
 configs:
 - config_name: default
   data_files:
@@ -19,4 +19,17 @@ configs:
     path: data/train-*
   - split: val
     path: data/val-*
+license: mit
+tags:
+- art
+size_categories:
+- n<1K
 ---
+
+# Dataset Summary
+Contains artwork with Cubism style.
+Used to finetune Stable Diffusion LoRA weights
+
+# Dataset Creation
+All data were scraped from the web (Google Image search)
+CLIP score was calculated with the prompt "a painting in the style of picasso's cubism" and the bottom 10% of the dataset was discarded.