Note: Below were experiments done at the early stage of this project.
See THIS PAGE
for more recent experiments where allowing rotation and flipping leads to cleaner templates.
Learning Active Basis Models from Non-aligned Images
The first plot is the learned template. For each of the following pairs of
images, the left is the observed image (displayed in color for visual clarity,
although the algorithm works on grey level images), the right is the image
superposed with the translated and deformed template.
Old code for reproducing results in paper:
Back to active basis homepage