Next: Input Encoding
Up: Neural Network Learning
Previous: Error Plots
- 20 people
- 32 images per person
- varying expression, direction looking, wearing sunglasses,
background, clothing worn, position of face in image
- 624 greyscale images, resolution 120x128, each pixel intensity
from 0 to 255
- could be many targets: identity, direction, gender, whether
wearing sunglasses - All can be learned to high accuracy
- we consider direction looking
Patricia Riddle
Fri May 15 13:00:36 NZST 1998