I have a problem: I just want to generate a TIFF format file with GPS information, can I not proceed with the 3D/2.5D modeling? In other words, is 3D/2.5D modeling necessary for picture fusion?
If you need a true orthorectified photo, yes! If you just need to stitch images together like the panorama app on your mobile phone, then no. Check out Microsoft ICE for that, combined wih QGIS for georeferencing.
If I only need to generate a GEOTIFF in ODM, how do I set it up (or modify the source code) to avoid unnecessary 3D reconstruction steps?
I’m afraid you can’t; ODM produces orthorectified imagery, and needs a 3D model to do so. You can speed things up a bit by passing
--skip-3dmodel, which avoids some computation. Also check
--fast-orthophoto for flat terrain imagery.