Skip to main content

Table 1 Pose prediction results of 10 noel views

From: Weakly-Supervised Single-view Dense 3D Point Cloud Reconstruction via Differentiable Renderer

Views Rotations (quarternion) Error (\(^\circ\))
View 1 Ours − 0.589 − 0.793 − 0.115 0.097 0.134
GT 0.589 0.793 0.113 − 0.096
View 2 Ours − 0.790 0.027 0.600 − 0.116 0.100
GT 0.790 − 0.027 − 0.599 0.117
View 3 Ours 0.331 − 0.175 − 0.422 0.824 0.210
GT 0.330 − 0.175 − 0.423 0.825
View 4 Ours − 0.587 − 0.375 − 0.263 0.666 0.269
GT 0.588 0.374 0.261 − 0.667
View 5 Ours − 0.556 − 0.320 − 0.435 − 0.630 0.340
GT 0.556 0.322 0.436 0.628
View 6 Ours 0.000 − 0.272 − 0.551 0.788 0.082
GT 0.001 − 0.271 − 0.551 0.788
View 7 Ours − 0.695 − 0.362 0.127 0.607 0.262
GT 0.695 0.361 − 0.129 − 0.607
View 8 Ours 0.242 0.657 0.293 0.650 0.234
GT 0.240 0.658 0.294 0.649
View 9 Ours − 0.511 − 0.470 0.422 0.581 0.197
GT 0.511 0.470 − 0.424 − 0.580
View 10 Ours − 0.746 − 0.217 − 0.048 − 0.627 0.288
GT 0.747 0.219 0.049 0.625