5. 1. User provides image
2. Preprocessing
• Automatic thresholding
• Scaling
• Centering
3. Generate Training Dataset
• Class 0: processed user image + variations
• Class 1…200: ‘R’s from random font subsample
6. 1. User provides image
Conv. Layer
Conv. Layer
Dense
Softmax
2. Preprocessing
• Automatic thresholding
• Scaling
• Centering
3. Generate Training Dataset
• Class 0: processed user image + variations
• Class 1…200: ‘R’s from random font subsample
4. Train Convolutional
Neural Network
7. 1. User provides image
Conv. Layer
Conv. Layer
Dense
Softmax
2. Preprocessing
• Automatic thresholding
• Scaling
• Centering
3. Generate Training Dataset
• Class 0: processed user image + variations
• Class 1…200: ‘R’s from random font subsample
4. Train Convolutional
Neural Network
5. Evaluate on all ‘R’s
in the font database
• Evaluate on all fonts
• Use output to order fonts
Python
Imaging
Library