First of all, thank you for developing this amazing package. It has really helped me handle stitching of very large datasets that don’t fit into memory.
Previously I used the BigStitcher plugin in FIJI. It works well, but when fusing multiple channels I often hit memory overhead, and the memory is not released properly before moving to the next channel (should call run("Collect Garbage") before fusing next channel dataset) . In addition, when running BigStitcher in FIJI headless mode, the macro doesn’t seem to return control back to the main process after the fusion is complete.
Recently I switched to BigStitcher-Spark, which works very well if I fuse all 4 channels at once. However, I run into problems when I try to fuse channels sequentially with different strategies. My workflow looks like this:
1) create container with 4 channels
create-fusion-container -ch 4
2) fuse channel 0 with AVG_BLEND (for structure / DAPI)
affine-fusion -f AVG_BLEND --timepointIndex 0 --channelIndex 0
3) fuse channel 1 with LOWEST_VIEWID_WINS (for spot structures)
affine-fusion -f LOWEST_VIEWID_WINS --timepointIndex 0 --channelIndex 1
4) fuse channel 2 similarly
affine-fusion -f LOWEST_VIEWID_WINS --timepointIndex 0 --channelIndex 2
The commands run, but the resulting dataset looks like the channel intensities are averaged together. I suspect this happens because I did not explicitly specify the ViewIds.
I am a bit confused about the logic here. If I want to fuse one channel with a different fusion strategy, do I really need to specify all the ViewIds for that channel (across angleId, illuminationId, tileId, etc.)? With large datasets there are dozens of ViewIds, which defeats the purpose of BigStitcher’s automation.
Question:
Is there a way to simply specify --channelIndex and let the other dimensions (ViewId, angleId, illuminationId) default to what is defined in the XML?
First of all, thank you for developing this amazing package. It has really helped me handle stitching of very large datasets that don’t fit into memory.
Previously I used the BigStitcher plugin in FIJI. It works well, but when fusing multiple channels I often hit memory overhead, and the memory is not released properly before moving to the next channel (should call
run("Collect Garbage")before fusing next channel dataset) . In addition, when running BigStitcher in FIJI headless mode, the macro doesn’t seem to return control back to the main process after the fusion is complete.Recently I switched to BigStitcher-Spark, which works very well if I fuse all 4 channels at once. However, I run into problems when I try to fuse channels sequentially with different strategies. My workflow looks like this:
The commands run, but the resulting dataset looks like the channel intensities are averaged together. I suspect this happens because I did not explicitly specify the ViewIds.
I am a bit confused about the logic here. If I want to fuse one channel with a different fusion strategy, do I really need to specify all the ViewIds for that channel (across angleId, illuminationId, tileId, etc.)? With large datasets there are dozens of ViewIds, which defeats the purpose of BigStitcher’s automation.
Question:
Is there a way to simply specify --channelIndex and let the other dimensions (ViewId, angleId, illuminationId) default to what is defined in the XML?