Object counting considers estimation of the number of specific objects in the
image. Solutions based on category-specific object detectors have been
extensively explored, however, they require huge annotated training datasets and
are not applicable to counting new, previously unobserved, classes. The latter
problem is explored by low-shot counting, which encompasses few-shot and
zero-shot counting. Few-shot counters count all present objects of some class
with only few of them annotated by bounding boxes, while zero-shot counters
consider counting the most frequentclass without annotations. In this talk I
will overview our recent counter LOCA that has been holding top performance on
the papers-with-code low-shot counting for over half of the year and will also
talk about a preliminary work, which exceeds that performance .