This is pretty neat! Small comment though:

The high model has ~400 vertices but ~1500 texture coordinate definitions. When you pass vertex data to OpenGL you must have as many vertices as texture coordinates because you may only use one index buffer per draw call. Based off of the simple texture that comes with it, is there any way to reexport this so that we don't use 1500 verts for a single traffic cone?

Thanks! I'm being told this model is being used in a media computer science AI course at HTW uni in Berlin, Germany. There's a licenses.txt file in the exercise which contains attribution.