meetran commited on
Commit
7a8a4f7
·
verified ·
1 Parent(s): c228313

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +68 -6
README.md CHANGED
@@ -1,13 +1,75 @@
1
  ---
2
- title: Painting Classifier Keras Final
3
- emoji: 📊
4
- colorFrom: blue
5
- colorTo: yellow
6
  sdk: gradio
7
- sdk_version: 5.49.1
8
  app_file: app.py
9
  pinned: false
10
  license: mit
11
  ---
12
 
13
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ title: WikiArt Painting Classifier
3
+ emoji: 🎨
4
+ colorFrom: purple
5
+ colorTo: blue
6
  sdk: gradio
7
+ sdk_version: 4.44.0
8
  app_file: app.py
9
  pinned: false
10
  license: mit
11
  ---
12
 
13
+ # WikiArt Painting Classifier - Interactive Demo
14
+
15
+ An interactive Gradio application for classifying paintings by artist, genre, and style.
16
+
17
+ ## Features
18
+
19
+ - **Multi-Task Classification**: Predicts artist, genre, and style simultaneously
20
+ - **129 Artists**: Including Monet, Van Gogh, Picasso, Da Vinci, Rembrandt, and more
21
+ - **11 Genres**: Portrait, landscape, abstract painting, still life, cityscape, etc.
22
+ - **27 Art Styles**: Impressionism, Cubism, Renaissance, Baroque, Expressionism, and more
23
+ - **Real-Time Predictions**: Instant results with confidence scores
24
+
25
+ ## How to Use
26
+
27
+ 1. Upload a painting image using the interface
28
+ 2. Click "Classify Painting" or wait for automatic classification
29
+ 3. View predictions for:
30
+ - Top 10 likely artists with confidence scores
31
+ - Genre classification with probabilities
32
+ - Top 10 likely art styles with confidence scores
33
+
34
+ ## Model Details
35
+
36
+ This app uses the model hosted at: [meetran/painting-classifier-keras-v1](https://huggingface.co/meetran/painting-classifier-keras-v1)
37
+
38
+ ### Architecture
39
+ - **Base Model**: MobileNetV2 (pre-trained on ImageNet)
40
+ - **Approach**: Multi-head classification with shared feature extraction
41
+ - **Framework**: TensorFlow/Keras
42
+ - **Input Size**: 224x224 RGB images
43
+
44
+ ### Training Details
45
+ - **Dataset**: WikiArt dataset with 84,440 paintings
46
+ - **Split**: 75% training, 25% validation
47
+ - **Augmentation**: Random horizontal flip, rotation, zoom
48
+ - **Training Stages**: Frozen backbone (2 epochs) + full fine-tuning (10 epochs)
49
+
50
+ ## Supported Classes
51
+
52
+ ### Artists (129 classes)
53
+ Claude Monet, Vincent van Gogh, Pablo Picasso, Leonardo da Vinci, Rembrandt, Salvador Dali, Edgar Degas, Pierre-Auguste Renoir, Paul Cezanne, Henri Matisse, Michelangelo, Raphael, Edvard Munch, Gustav Klimt, Wassily Kandinsky, and 114 more.
54
+
55
+ ### Genres (11 classes)
56
+ Abstract painting, Cityscape, Genre painting, Illustration, Landscape, Nude painting, Portrait, Religious painting, Sketch and study, Still life, Unknown Genre
57
+
58
+ ### Art Styles (27 classes)
59
+ Impressionism, Post-Impressionism, Cubism, Abstract Expressionism, Early Renaissance, High Renaissance, Northern Renaissance, Mannerism Late Renaissance, Baroque, Rococo, Romanticism, Realism, Expressionism, Symbolism, Art Nouveau, Fauvism, Pointillism, Naive Art Primitivism, Synthetic Cubism, Analytical Cubism, Action painting, Color Field Painting, Contemporary Realism, New Realism, Minimalism, Pop Art, Ukiyo-e
60
+
61
+ ## Limitations
62
+
63
+ - Best performance on Western classical and modern art from the training distribution
64
+ - May have reduced accuracy on contemporary art, digital art, or non-Western styles
65
+ - "Unknown Artist" and "Unknown Genre" classes may have lower precision
66
+ - Requires clear, well-lit images for optimal results
67
+
68
+ ## License
69
+
70
+ MIT License
71
+
72
+ ## Links
73
+
74
+ - **Model Repository**: https://huggingface.co/meetran/painting-classifier-keras-v1
75
+ - **Training Dataset**: [WikiArt on Hugging Face](https://huggingface.co/datasets/huggan/wikiart)