Anonymous
Learning based PatchMatch MVS. CVPR 2022 submission
Model ran with:
- image resolution of 1920x1072
- 15 Source Images
- 3 View selections per pixel
- Fused with relative inlier scale of 0.01, projection distance of 2px, and anglular difference of 10' with at least 5 views (including Ref)
Precision | Recall |