Species detections often vary depending on the survey methods employed. Some species may go undetected when using only one approach in community-level inventory and monitoring programs, which has management and conservation implications. We conducted a comparative study of terrestrial mammal and bird detections in the spring and summer of 2021 by placing camera traps at 30 locations across a large military installation in northern Michigan, USA and testing replicate soil samples from these sites for environmental DNA (eDNA) using an established vertebrate metabarcoding assay. We detected a total of 48 taxa from both survey methods: 26 mammalian taxa (excluding humans, 24 to species and two to genus) and 22 avian taxa (21 to species and one to genus). We detected a relatively even distribution of mammalian taxa on cameras (17) and via eDNA analysis (15), with seven taxa detected from both methods. Most medium-to-large carnivores were detected only on cameras, whereas semi-fossorial small mammals were detected only via eDNA analysis. We detected higher bird diversity with camera traps (18 taxa) compared to eDNA analysis (eight taxa; four taxa were detected with both methods), but cameras alone were most effective at detecting smaller birds that frequently occupy arboreal environments. We also used Bayesian spatial occupancy models for two widely distributed game species (white-tailed deer, Odocoileus virginianus, and ruffed grouse, Bonasa umbellus) that were moderately detected with both survey methods and found species-specific site use (occupancy) estimates were similar between cameras and eDNA analysis. Concordant with similar studies, our findings suggest that a combination of camera trap and eDNA surveys could be most useful for assessing the composition of terrestrial mammal communities. Camera traps may be most efficient for assessing bird diversity but can be complemented with eDNA analysis, particularly for species that spend considerable time on the ground.