Relatively few studies have directly examined children’s memory of object-based spatial structure of room-sized environments. The current study investigated how children remember the spatial structure of a room, and the role of pictorial working memory (WM) and different testing perspectives in this process. In Experiment 1, 80 children aged 5 to 7 years old participated in a memory of spatial structure task, a pictorial spatial WM task, and a verbal ability test. In the first task, participants explored a square-shaped virtual room where each wall was associated with different objects. The virtual room was viewed on a desktop monitor. Then they recalled wall scenes of different perspectives (0, 90, 180 degrees) relative to a reference wall. Results found that pictorial WM predicted memory of spatial structure after accounting for age and verbal ability. Different from perspective taking tasks, children rarely made egocentric errors. Moreover, the memory of spatial structure is perspective independent because there was no difference in performance between the 90 and 180 degrees conditions. We replicated the basic results in Experiment 2, where 77 children completed the same set of tasks except that the virtual room was round-shaped. Furthermore, the developmental trajectories of memory of spatial structure as a function of pictorial WM varied between different experiments and perspective conditions. Together, our study showed that children aged 5–7 years old could retrieve the spatial structure of a room-sized environment from diverse perspectives, which was also susceptible to pictorial WM capacity, perspectives, and geometric cues.