BrowserQuest map in question.

From what I understand, with tiled, there are two main ways to specify collision:

1. Create an object layer, and interpret the shapes in the engine as collision objects.
2. Create a tiled layer, and make all tiles in the layer have a collision property, and interpret all tiles in the layer as collision objects.

I'm using BrowserQuest as a big source of inspiration for my project, and I want to know how they handled collision on the level editing side.

I've checked through all their layers, expecting an object layer to be handling cliff collision like:

But there are no such object layers to be found. Furthermore, the tile layers containing the tiles for such cliffs have no properties at all, meaning that they didn't just specify "collision" for such tile layers.

I especially need to know how they handled less rectangular shapes like:

I could imagine that they are not using explicit collision layers, but instead determining collision in the actual engine, based off the presence of specific tile layer sprites. Only because BrowserQuest has whole-tile movement, and it wouldn't look too odd if a small apple, taking up only a fraction of the tile size, prevents movement over that entire tile.

But I'm creating a game with more precise movement, so collision has to be tight to the apple, and I really want to know how BrowserQuest approached collision defining.

If anyone knowledgeable with Tiled could take a quick look at the map, I'd appreciate it! I'm tearing my hair out here :). Thanks