ShapeMatcher Step-By-Step Tutorial

1. Download and unzip it in c:\

2. Download and unzip it in c:\ShapeMatcher

3. Create a shape database with the 3 images in the ShapeMatcher folder

4. Match the second image (ID 1) against all objects in the database (ignore warnings)

5. Type 'y' to see the first node correspondence results for the first match. The title bar of the window shows
which keys you can type (invalid keys close the window). In addition, 'c' would turn on/off the coordinates, and
numbers from 0 to 9 would turn on/off the correspondence lines: 0 = show all lines, 9 = only lines with 0.9 - 1
similarity, 8 = 0.8 - 0.9 similarity, etc.

6. Close the window to see the next match by typing 'q' or the spacebar with the focus on the
node correspondence window.

7. The next match is more interesting since it would be the top ranqued shape if the "query" shape wasn't in the database.

8. Type 'a' to see the shock graph of the "query" shape [the shape on the left]

9. Type 'b' to see the shock graph of the "model" shape [the shape on the right]

10. To see a mode detailed summary of the results, open the first_results.log file with a text editor. This files shows indexing and
matching resuls along with the parameter values used during the computation.

11. Now, try building more interesting query and model databases using the images in the Model120 folder. First,
build the query database using noisy images. Second, build the model database using clean images. That is,

> sm -c query.db Models120\Bumps5a

> sm -c model.db Models120\Clean

If you want this to be done much faster, turn off the external simplification step. Eg,

> sm -doExtSimp 0 -c query.db Models120\Bumps5a

> sm -doExtSimp 0 -c model.db Models120\Clean

12. You can now match each object in the query database against all object in the model database by

13. Note that the left image has "bumps" on its boundary that are not present on the right image.

14. The next best match is a different viewpoint of the "model" bull.