Supplementary Material for WT-MVSNet: Window-based Transformers for Multi-view Stereo Jinli Liao 1,2 Yikang Ding

As shown in Tab. 2, we explore the influence of different number of attention blocks With the expansion of window size, the inference time reduces. WET on DTU evaluation set [2] ( lower is better). DTU evaluation set [2], and the results of point clouds are summarized in Tab. 5. 2 Table 5: Ablation study on the number of input images Our model demonstrates its excellent robustness and generalizability in a variety of scenarios.