360SD-Net: 360° Stereo Depth Estimation with Learnable Cost Volume: Difference between revisions

360SD-Net: 360° Stereo Depth Estimation with Learnable Cost Volume (view source)

381 bytes removed , 15 June 2020

5,337

edits

@@ Line 7: / Line 7: @@
 ==Method==
-Their input is two equirectangular images, one taken above another.
-They also input the polar angle:
-<pre>
-# Y angle
-angle_y = np.array([(i-0.5)/512*180 for i in range(256, -256, -1)])
-angle_ys = np.tile(angle_y[:, np.newaxis, np.newaxis], (1,1024, 1))
-equi_info = angle_ys
-</pre>
-The angles are equivalent to <code>np.linspace(90, -90, height+1)[:-1] - 0.5*(180/height)</code>
 ==Architecture==