Estimating duiker populations is essential for conservation programmes in Central Africa, but what the most reliable survey method is remains unclear. We estimated forest duiker population densities in Southeast Cameroon using three line‐transect methods: dung counts and diurnal direct observation over 338.2 km, and nocturnal direct observation over 334.7 km. Nocturnal observations resulted in 59.8 and 9.0 individuals km−2 for the blue duiker (Philantomba monticola) and red duikers (Cephalophus callipygus, C. dorsalis, C. leucogaster and C. nigrifrons), respectively, which were much higher than estimates by dung counts and diurnal observation. Only dung counting was applicable for the yellow‐backed duiker (0.8 individuals km−2), due to low encounter rates obtained by the other methods. We argue that nocturnal observation is the most accurate method because the torchlight can pick out duikers' eyes even if the animals are shy and hiding and because multiple counts are unlikely. By contrast, the other methods probably underestimate densities due to poor visibility and reliance on dung production and decay rates calculated in different areas. These results can affect figures used in wildlife management decisions, such as the maximum sustainable yield of game animals, which generally have been based on densities estimated using a single survey method.