Mammal inventories in tropical forests are often difficult to carry out, and many elusive species are missed or only reported from interviews with local people. Camera traps offer a new tool for conducting inventories of large- and medium-sized terrestrial mammals. We evaluated the efficiency of camera traps based on data from two surveys carried out at a single site during 2 consecutive years. The survey efforts were 1440 and 2340 camera days, and 75 and 86% of the 28 large- and medium-sized terrestrial mammal species known to occur at the site were recorded. Capture frequencies for different species were highly correlated between the surveys, and the capture probability for animals that passed in front of the cameras decreased with decreasing size of the species. Camera spacing and total survey area had little influence on the number of species recorded, with survey effort being the main factor determining the number of recorded species. Using a model we demonstrated the exponential increase in survey effort required to record the most elusive species. We evaluated the performance of different species richness estimators on this dataset and found the Jackknife estimators generally to perform best. We give recommendations on how to increase efficiency of camera trap surveys exclusively targeted at species inventories.