Hello cryoSPARC users!
We’re working on a significant overhaul of the exposure curation tool. Many of you have expressed that while the tool is useful, there are still areas to improve on. We have made a list of suggestions and feature requests (as well as bugs to mitigate) based on all the other topics created, but I’d like to open this topic so we can keep track of future suggestions within a single one.
Feedback is welcome 
Regards,
Suhail
Hi Suhail,
Looking forward to seeing the new version - the exposure curation tool has a lot of potential!
I have many specific thoughts, but in general the UI needs to be greatly simplified, and optimized for larger datasets - currently it gets incredibly laggy and crash-prone with datasets of several thousand mics.
The little toolbar up the top of the right hand panel is very confusing and IMO mostly unnecessary - every time I use it I have to think about whether I want to accept/reject all, accept/reject selection, save, or “save and done” (?). The utility of these different options is not clearly apparent (why would I want to “accept all” when I am curating exposures?).
The text entry boxes for parameters should work - currently the user can enter values but nothing happens - only the sliders work. This makes it hard to adjust parameters precisely, particularly as the sliders themselves are quite laggy/jerky with large datasets.
Also, the drop down for selecting parameters to plot in the graph should show the full name of each parameter - currently it only shows the first several characters, which makes it very hard to determine what I am picking (without hovering to wait for a tooltip to appear).
The parameter graph also takes up a lot more screen real estate than it needs to, in my opinion - I wonder whether it would be possible to show both the overall and indivdual views side by side, to allow easy exploration of parameters - e.g. clicking on mics with different CTF fit res or motion correction stats and seeing the mic/CTF on the right hand side would be great. I would also simplify the zooming/manipulation tools in the param graph - adjustment of axes is all one really needs here.
I also wonder if maybe, given the number of parameters that are now present, it would be better to only show adjustment sliders/boxes for the parameters that are currently plotted, rather than showing all the undisplayed params also.
Cheers
Oli
Nice. I’m glad to see the developers so interested in user feedback. I know it has got to be hard to implement a million little user experience features, but it really is appreciated.
Anyway, since we’re on the topic, I think a few features could be added to the ‘inspect particle picks’ tool. As is, it can be a bit laggy when adjusting the two sliders for thresholds. I think this is because it recalculates the number of particle picks on the fly for several thousand micrographs. This means any time the slider changes it freezes for like a minute or two as it recalculates. Maybe the sliders effect only the image displayed, then you can click a button like “apply to remaining micrographs” or something. Also I notice that those sliders only ever can give you less picks than the auto picker found, but is it possible to slide them in a way to increase the number of picks? Maybe not necessary, just a thought.
thanks again!
I would also like to be able to switch between the two-parameter graph view, and a histogram view - sometimes it is useful to be able to view a histogram of e.g. defocus or astigmatism, to get a sense of the distribution of the data.
Cheers
Oli
I’d also like to be able to curate based on the local defocus variation - I think a non-physical defocus variation is an easy parameter to use for rejection. Likewise, sorting by relative ice thickness would also be useful.
Cheers
Oli
One more thought - the exposure curation tool, apart from being useful for curating exposures, is also useful as just a viewer, for exploring the dataset. Unfortunately, once you click done, you can no longer use it for that purpose. Would it be possible to alter it such that after submission the user can no longer submit, but can still interact with the dataset and view micrographs? The individual view with the CTF and morion correction info, as well as the micrograph, is particularly useful.
Text entry boxes work if you type the number, then click outside the box, then hit tab. Bug.
I agree a simpler way to cull the data would be useful. For instance, extreme outliers ruin the scale of the axes and thus the usefulness of the tool. I typically have to curate first to remove extreme outliers of all parameters, create the outputs, and relaunch the tool on the already curated dataset.
Yes I have also encountered this - e.g. displaying the CTF cross correlation data is usually useless because one outlier will throw it off. I wonder if one way to go would be to rescale the axes after accepting a selection - that should alleviate the outliers issue a bit.
Hi Suhail - one more suggestion in the rework of this tool - could you please write out the thresholds that were used to the log file? This is useful for record keeping and writing methods later - e.g. “all micrographs with CTF fit worse than x Å or defocus greater than y µm were rejected” etc
Inspect picks. writes out the thresholds used, but exposure curation currently does not.
Cheers
Oli
I like the idea of curating exposures, including by average intensity. Wouldn’t it be trivial to parallelize
with cpus to speed up the average value part? It does warn that it takes a long time, but for a large data set it is really long time.
Thanks for all your efforts! George
When using larger datasets (e.g. 13000 mics) the curate micrographs tool is entirely unusable - any prospects for speeding it up in the overhaul? Is there a timeline for the new version?
Hi @sdawood - is there any timeline for the overhaul of Curate Micrographs? It is becoming increasingly unusable as typical dataset sizes have increased with wider availability of K3 detectors - e.g with 15k micrographs it is really not practical. Even just addition of non-interactive thresholds (as exist now for Inspect Particles) would make a big difference! Hopefully in the next version?
Hi everyone,
Now that v3 and the newly redesigned Live application have been released, we are focused on updating all interactive jobs. This includes the significant changes to exposure curation that has been in the works for some time now. Please keep an eye out for updates in the next few weeks!
- Suhail
Hi Suhail,
I’m wondering if I can have the list of all accepted images’ names in the Curate Exposure Job? As I’m going to subtract gold beads from my micrographs by using IMOD software automatically, before that, if I can first trash all the un-accepted images in the Curate Exposure Job, it might be more efficient for the later steps.
Best,
Chuchu
Hi,
It would be nice to have a quick way to see the selection/rejection criteria (like selected images or thresholds for selected images) that were used in a previous job of Exposure Curation e.g. in the input/output mask.
Best,
christian
@sdawood, using version 3.2 of Cryosparc and currently frustrated by a number of bugs with manual curation. Sliders update slowly and inputting a value seems to sometimes apply that value to another parameter (ie. changing ice thickness seems to apply changes to eg. total motion). Is the overhaul of this feature still in train?
It would be nice to see the cutoffs that were applied in the curate exposure log after the job has been completed, sometimes I don’t keep track and don’t remember what cutoffs I actually applied
In addition, can we have inputs in the job to specify the cutoffs automatically? I usually use similar values anyways, so often there is no reason for me to actually look at the micrographs.
Hi @sdawood - is this still in the works? Even a small, temporary update would be appreciated, e.g. adding non-interactive thresholds like for inspect picks & outputting threshold values to the log, if the more substantial rework will still take some time?
It would be great if exposure curation generated a summary of statistics in the log file (or maybe as a pdf attached to the log file. E.g. a histogram of each parameter (or at least defocus, relative ice thickness and CTF fit res), with mean, min and max values. This would be useful for record keeping and adjusting downstream processing (e.g. the average defocus is useful to know in guiding what box size to use).