Camera-trap surveys, when combined with capture-recapture models, are an accurate and cost-effective method for estimating the abundance of individually identifiable carnivore species such as leopards, Panthera pardus. Reliable population estimates for leopards are particularly useful as they enable informed conservation and management decisions for a species that, although widespread, is heavily persecuted. Here we present the results of a camera-trap survey that estimated the population density of leopards in Zululand Rhino Reserve (ZRR) in northern KwaZulu-Natal, South Africa. We divided our study area in half and subsampled each section for 40 days using 34 paired camera-trap stations. The combined data yielded captures of six individual leopards (3 male, 3 female) photographed on 19 occasions. Using the program CAPTURE, population abundance was estimated according to two models; one that assumes homogeneous capture probabilities among individual leopards (M_o) and one that allows for heterogeneity in capture probabilities between individuals (M_h). For model M_o, the resultant capture probability was 0.317 and estimated abundance was 6 ± 0.390 (6–6, 95% Cl). For model M_h, the resultant capture probability was 0.112 and estimated abundance was 17 ± 6.37 (10–37, 95% Cl). The area effectively sampled by the camera-traps was 234 km². Therefore, the estimated population density of leopards in the reserve was 2.5 ± 0.195 leopards per 100 km² according to model M_o, and 7.00 ± 2.64 leopards per 100 km² according to model M_h. When compared with the results of similar surveys conducted in nearby reserves, our findings suggest that the leopard population in the ZRR is below carrying capacity. To encourage population recovery, we recommend that no trophy hunting of leopards be conducted in the reserve, that measures be introduced to reduce retaliatory killing of leopards by livestock owners in surrounding areas, and that the illegal trade in leopard skins be addressed.