- We parse the input and output scenes into objects.
- Avoiding unwanted object globulation:
- When parsing the output scene of example 1, two of the objects are touching diagonally, which currently causes our system to form a single object rather than two. This is problematic.
- Likewise, when parsing the test output scene, three of the objects are touching, causing them to form a single object rather than three. This also is problematic.
- Strategy
- Parse the input and output scenes and produce a library of notable objects that we’ve seen.
- For each object in each scene, check whether the object could be reinterpreted as multiple objects from our library of notable objects.
- If so, break the object up.
For each input object, there is only one object in the output scene that it is contained within, making it straightforward to produce a mapping from input objects to output objects.
e.g.
All object mappings involve replacing the input object with a common output image, making it straightforward to produce a single rule:
