Measuring the flow field not just above but also inside the canopy of rough beds in open-channel flows is challenging when optical access from below or the side is not possible. To this end, a new stereo-PIV arrangement was implemented and tested. To optically access the area in between the roughness elements, a top-viewing stereo-PIV system was installed at a steep viewing angle of 70∘\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$^{\\circ }$$\\end{document}. Furthermore, a glass plate was installed at the water surface to avoid random image distortions from surface waves. Finally, in-house-produced low-cost fluorescent particles were used to filter out laser reflections on the canopy’s surface. To validate the stereo-PIV measurements with respect to the steep viewing angle of the stereo cameras, the instantaneous flow field above a bed of spheres was measured simultaneously with the stereo-PIV system and with a third side-looking camera for standard 2D-PIV processing. Two different water depths and two different Reynolds numbers were investigated under uniform flow conditions. Finally, to assess the influence of the glass plate on the flow, 2D-PIV measurements were performed with and without the glass plate. It is shown that the steep viewing stereo PIV can properly reconstruct the instantaneous, mean and turbulent statistics except for the vertical normal turbulent stresses, which are overestimated due to peak-locking errors. Also, the boundary layer developing below the glass plate induces a slight acceleration in the bulk flow, which can be considered negligible for higher water depths. In all instances, the flow acceleration does not affect the near-bed region.Graphical abstract