We were stitching the datasets for all the channels using the displacements that were calculated for one of the channels. Now, this is not optimal since we're doing 3D stitching and we are running out of memory.
Possible solution outline:
[x] 1. Stitch channels independently using the calculated displacements for the informative channel.
[x] 2. Calculate the needed memory to perform parallel stitching of multiple channels. If the needed memory is greater than the available memory, then we need to stitch the channels separately. This would also depend in the degree of parallelization, if needed we might need to find the optimal degree of parallelization.
[x] 3. Get data about computation time, dataset size, # tiles, grid size, etc., per step for version 1.0, version 2.0 with multiple channel stitching at a time and version 2.0 with independent channel stitching.
We were stitching the datasets for all the channels using the displacements that were calculated for one of the channels. Now, this is not optimal since we're doing 3D stitching and we are running out of memory.
Possible solution outline:
version 1.0
,version 2.0 with multiple channel stitching at a time
andversion 2.0 with independent channel stitching
.