File size: 1,528 Bytes
809650c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 |
---
license: MIT
language: en
license: mit
datasets:
- web crawled (coming soon)
---
# Simple CNN-based Artist Classifier
This repo contains a simple CNN-based Keras model which classifies images into one of 10 selected artists/painters.
- The purpose of this model was for a quick prototyping
- Data has been web-crawled using `https://github.com/YoongiKim/AutoCrawler`
- 10 popular artists/painters were chosen:
- \[ARTIST\]: \[ID\]
- claude_monet: 0,
- henri_matisse: 1,
- jean_michel_basquiat: 2,
- keith_haring: 3,
- pablo_picasso: 4,
- pierre_augste_renoir: 5,
- rene_magritte: 6,
- roy_richtenstein: 7,
- vincent_van_gogh: 8,
- wassily_kandinsky: 9
- About 100 representative paintings per artist were crawled and manually checked
- Dataset will be shared later
# How to use
```python
import tensorflow as tf
from huggingface_hub import from_pretrained_keras
model = from_pretrained_keras("jkang/drawing-artist-classifier")
image_file = 'cat.jpg'
img = tf.io.read_file(image_file)
img = tf.io.decode_jpeg(img, channels=3)
last_layer_activation, predictions = model(img[tf.newaxis,...])
```
# Intended uses & limitations
You can use this model freely for predicting artists or trends of a given image.
Please keep in mind that this model is not intended for the production, but for a research and quick prototyping.
Web-crawled image data might not have balanced amount of drawings that sufficiently represent the artists.
---
- 2022-01-18 first created by jaekoo kang |