Gens 1-3 are all tile based, however ever since gen 4, pokemon hasn't been tile-based. Instead it started using 2.5d for gen 4-5 with 3d objects. With that being said, both of these gens can appear tile based still. Bulbapedia actually stores images of all of the cities, routes, dungeons, and locations in each game. If we can pull the tiles from an image, and then translate this to mapgrid and metatiles in gen 3, we can export every map for gen 4 and 5 to complete sinnoh and unova and save us lots of time.
Design
I believe we can use a python script using some computer vision libraries to parse an image given a "slicing" square, and then count every tile that gets used across the different maps, and increments on the same tiles. Now the games aren't tile based, so the images aren't going to be exactly one to one in pixels, this shouldn't be a problem. Instead we can have a threshold for image "likeness" to identify similar tiles that get used across the project. We'd then get a list of most used tiles, and then create a mapping from the sinnoh tiles, to tilesets and metatiles in gen 3. This could automate a lot of the grunt work
Description
Motivation
Gens 1-3 are all tile based, however ever since gen 4, pokemon hasn't been tile-based. Instead it started using 2.5d for gen 4-5 with 3d objects. With that being said, both of these gens can appear tile based still. Bulbapedia actually stores images of all of the cities, routes, dungeons, and locations in each game. If we can pull the tiles from an image, and then translate this to mapgrid and metatiles in gen 3, we can export every map for gen 4 and 5 to complete sinnoh and unova and save us lots of time.
Design
I believe we can use a python script using some computer vision libraries to parse an image given a "slicing" square, and then count every tile that gets used across the different maps, and increments on the same tiles. Now the games aren't tile based, so the images aren't going to be exactly one to one in pixels, this shouldn't be a problem. Instead we can have a threshold for image "likeness" to identify similar tiles that get used across the project. We'd then get a list of most used tiles, and then create a mapping from the sinnoh tiles, to tilesets and metatiles in gen 3. This could automate a lot of the grunt work
Discord contact info
No response