Ruofan Zhou is asking:
"
We have two questions regarding the Bokeh Effect Challenge, we would be very appreciate if you could answer them:
(1)
Is it allowed to use external data or external pre-trained models? We were thinking of using some depth-estimation networks, either use the pre-trained model or we train it on some external dataset, but we are not sure if it is allowed to do so.
(2)
We notice there are some alignment artifacts in some images (for example shifts in the edges, 0-value pixels along the boundary). Thus we want to ask would you also release the non-aligned "Bokeh images" or the details of the alignment algorithm you've used?
"
Response:
(1)
Yes, you can use external data, pretrained models, or other tools, (for depth-estimation, segmentation, etc..) The sole condition is to clearly mention this in your submission / solution factsheet.
(2)
No, we do not plan to release any other data / algorithms during this challenge, maybe after.