2017-09-05

Sample mixing layer, rotation invariance and AA

Rotation layer: performance is a problem
Rotation layer: no clear improvement over standard convolutions (AA)

Average MSE over 30 images, 1spp reconstruction.

pixel no angle ours fc
Edges 0.000289 0.000374 0.000237 --
Sines 0.000340 0.000330 0.000225 --
Triangles 0.000733 0.000847 0.001081 0.000924

Edges ref|1spp|pixels|noangle|ours Edges ref|1spp|pixels|noangle|ours

Sines ref|1spp|pixels|noangle|ours Sines ref|1spp|pixels|noangle|ours

Triangles ref|1spp|pixels|noangle|ours|fc Triangles ref|1spp|pixels|noangle|ours|fc

Rotation layer: other concerns
Other option? Learn pairwise sample interactions directly