Closed whattheweather closed 4 years ago
@whattheweather
Forward of the network, clustering, and recovering planar depth were included.
@niujinshuchong Thanks, so I time from
# forward pass
logit, embedding, _, _, param = network(image)
(https://github.com/svip-lab/PlanarReconstruction/blob/master/main.py#L415) to
# infer instance depth
instance_loss, instance_depth, instance_abs_disntace, instance_parameter = \
instance_parameter_loss(segmentation, sampled_segmentation, sample_param,
valid_region, gt_depth, False)
(https://github.com/svip-lab/PlanarReconstruction/blob/master/main.py#L436) , is that what you mean?
Or should I time until
# mask out non planar region
predict_segmentation[prob.cpu().numpy().reshape(-1) <= 0.1] = 20
predict_segmentation = predict_segmentation.reshape(192, 256)
(https://github.com/svip-lab/PlanarReconstruction/blob/master/main.py#L459) ?
@whattheweather The first one.
@whattheweather The first one.
Got it! Thanks!
Hi,
In the paper, you got 32.26FPS on Titan XP, and which processes were included when you calculated it?
Thanks!