On the Limits of Pseudo Ground Truth in Visual Camera Re-Localization —Supplementary Material—

Sec. 1 presents implementation details for our baseline methods (c.f . Sec. 5 “Baselines” in the main paper). Sec. 2 provides visualisations comparing the original and the Structure-from-Motion (SfM) pseudo ground truth (pGT), complementing Figures 1, 3, 5, and 6 in the main paper. Sec. 3 extends Figure 4 in the main paper by providing 3D alignment statistics for all scenes in the 7Scenes [16] and 12Scenes [18] datasets. Finally, Sec. 4 provides detailed plots for the localization results presented in Sec. 5 of the main paper.