Both executables rely on trained vector quantization prototypes. Such data can be trained with the procedure described in Training Stage. Alternatively, pre-trained data can be downloaded from Polyora's web site.
Please note that the pre-trained data might not work well with your configuration. Training your own data is more reliable.
../qtpolyora/qtpolyora -rp -d descriptors.dat -v visual.db
../buildtree/buildtree -d descriptors.dat -r 8 -e 64 -v visual.db
../qtpolyora/qtpolyora -rt -d tracks.dat -v visual.db
../cluster_ids/cluster_ids -d tracks.dat -T .1 -v visual.db
The resulting visual.db should contain the quantization tree for descriptors, and the prototypes for quantizing tracks. Visual objects can also be stored in the same file.
If you have large files (>2 GB), compiling and running these tools in a 64 bits environment might be necessary. Please note also that cluster_ids can require a large amount of memory.
vobj_tracker -v visual.db