Recent advances in photogrammetry and computer vision have seen a growing interest in 3D image-based modelling. Semi-automatic workflows for processing large image sets are now available as commercial and open-source solutions, and coupled with low cost photo-imaging sensors, a new era of slope visualisation opportunities has begun. This paper investigates the application of low-cost image sensors, including a smartphone, two off-the-shelf compact digital cameras and a Raspberry Pi camera module, and low-cost aerial platforms, including one quadcopter with an action camera and another quadcopter with a built-in camera, to geotechnical problems. The study showed that reliable rock surface models can be created from images acquired by any of the different sensors tested, with the photogrammetric models able to match a laser scanner model to within 5 to 11 mm for the majority of areas on the slope. Reliability decreased for areas that were in heavy shade at the time of image acquisition, with errors of as much as 28 mm in extreme cases, although these were unusual. Models derived from terrestrial and aerially-mounted sensors performed similarly well, except for images acquired by a low-cost UAV for which the only positional control was an on-board GPS, which could only achieve an accuracy of 80-150 mm for the majority of points. Dips inferred from features in the virtual model were generally accurate to within 1 to 2 degrees. The variation in the dip direction is bigger, and in some cases even more than 12 degrees. Several of the steeper virtual measurements also have reversed dip directions. The reversed virtual dip measurements were a numerical anomaly, associated with the interpretation of very steep dip angle values from the model.