inverted normals

Hi, if it wasn’t a bug, recomputing normals with meshlab wouldn’t fix it. I think it’s related to obj format