We determine the morphological mix of field galaxies down to mr similar or equal to 24.25 mag (m(B) similar to 26.0 mag) from a single ultradeep Hubble Space Telescope wide field planetary camera (WFPC2) image in both the V-606, and the I-814 filters. In total, we find 227 objects with m(I) less than or equal to 24.5 mag and classify these into three types: ellipticals (16%), early-type spirals (37%), and late-type spirals/irregulars (47%). The differential number counts for each type are compared with simple models in a standard flat cosmology. We find that both the elliptical and the early-type spiral number counts are well described by little-or-no-evolution models, but only when normalized at b(J) = 18.0 mag. Given the uncertainties in the luminosity function (LF) normalization, both populations are consistent with a mild evolutionary scenario based on a normal/low rate of star formation. This constrains the end of the last major star formation epoch in the giant galaxy populations to z greater than or equal to 0.8. Conversely, the density of the observed late-type/irregular population is found to be a factor of 10 in excess of the conventional no-evolution model. This large population might be explained by a modified local dwarf-rich LF and/or strong evolution acting on the local LF. For the dwarf-rich case, a steep faint-end Schechter slope (alpha similar or equal to -1.8) is required, plus a fivefold increase in the dwarf normalization. For a purely evolving model based on a flat Loveday et al. LF (alpha similar or equal to -1.0), a ubiquitous starburst of Delta similar to 2.0 mag is needed at z similar or equal to 0.5 for the entire late-type population. We argue for a combination of these possibilities, and show that for a steep Marzke et al. LF (alpha similar or equal to -1.5) a starburst of similar to 1.3 mag is required at z similar or equal to 0.5 in the entire late-type population, or similar to 2.0 mag in similar to 20% of the population.