The 3D magnetic field topology of solar filaments/prominences is strongly debated, because it is not directly measureable in the corona. Among various prominence models, several are consistent with many observations, but their related topologies are very different. We conduct observations to address this paradigm. We measure the photospheric vector magnetic field in several small flux concentrations surrounding a filament observed far from disc center. Our objective is to test for the presence/absence of magnetic dips around/below the filament body/barb, which is a strong constraint on prominence models, and that is still untested by observations. Our observations are performed with the THEMIS/MTR instrument. The four Stokes parameters are extracted, from which the vector magnetic fields are calculated using a PCA inversion. The resulting vector fields are then deprojected onto the photospheric plane. The 180° ambiguity is then solved by selecting the only solution that matches filament chirality rules. Considering the weakness of the resulting magnetic fields, a careful analysis of the inversion procedure and its error bars was performed, to avoid over-interpretation of noisy or ambiguous Stokes profiles. Thanks to the simultaneous multi-wavelength THEMIS observations, the vector field maps are coaligned with the Hα image of the filament. By definition, photospheric dips are identifiable where the horizontal component of the magnetic field points from a negative toward a positive polarity. Among six bipolar regions analyzed in the filament channel, four at least display photospheric magnetic dips, i.e. bald patches. For barbs, the topology of the endpoint is that of a bald patch located next to a parasitic polarity, not of an arcade pointing within the polarity. The observed magnetic field topology in the photosphere tends to support models of prominence based on magnetic dips located within weakly twisted flux tubes. Their underlying and lateral extensions form photospheric dips both within the channel and below barbs.