jkang commited on
Commit
8e394d8
1 Parent(s): d2d2315

upload readme

Browse files
Files changed (3) hide show
  1. .gitignore +1 -0
  2. .gitignore~ +0 -0
  3. model_card.md +50 -0
.gitignore ADDED
@@ -0,0 +1 @@
 
 
1
+ .ipynb_checkpoints
.gitignore~ ADDED
File without changes
model_card.md ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ license: mit
4
+ datasets:
5
+ - web crawled (coming soon)
6
+ ---
7
+
8
+ # Simple CNN-based Artist Classifier
9
+
10
+ This repo contains a simple CNN-based Keras model which classifies images into one of 8 artistic trends.
11
+
12
+ See also: `https://huggingface.co/jkang/drawing-artist-classifier`
13
+
14
+ - The purpose of this model was for a quick prototyping
15
+ - Data has been web-crawled using `https://github.com/YoongiKim/AutoCrawler`
16
+ - 8 popular artists/painters were chosen:
17
+ - \[TREND\]: \[ID\]
18
+ - cubism: 0,
19
+ - expressionism: 1,
20
+ - fauvisme: 2,
21
+ - graffitiar: 3,
22
+ - impressionism: 4,
23
+ - popart: 5,
24
+ - post_impressionism: 6,
25
+ - surrealism: 7}
26
+ - About 100 representative paintings per artist considering 8 trends were crawled and manually checked
27
+ - Dataset will be shared later
28
+
29
+ # How to use
30
+
31
+ ```python
32
+ import tensorflow as tf
33
+ from huggingface_hub import from_pretrained_keras
34
+ model = from_pretrained_keras("jkang/drawing-artistic-trend-classifier")
35
+
36
+ image_file = 'kandinski.jpg'
37
+ img = tf.io.read_file(image_file)
38
+ img = tf.io.decode_jpeg(img, channels=3)
39
+
40
+ last_layer_activation, predictions = model(img[tf.newaxis,...])
41
+ ```
42
+
43
+ # Intended uses & limitations
44
+ You can use this model freely for predicting artists or trends of a given image.
45
+ Please keep in mind that this model is not intended for the production, but for a research and quick prototyping.
46
+ Web-crawled image data might not have balanced amount of drawings that sufficiently represent the artists.
47
+
48
+
49
+ ---
50
+ - 2022-01-18 first created by jaekoo kang