awedwards / daphnia Goto Github PK

If the last image in a list isn't found (Such as in block xfo), the gui crashes to a grey screen. We aren't able to save or go back.

Something weird is also going on with the shape output. It repeats one of them numerous times (i.e., line 2,800 to line 34,000). Both of these errors are in the outputs labelled xadtest_20190108.

Theres also something weird going on with the masked regions output again. This screenshot was from when I was trying to figure something out;

And this is from a few minutes later;

In the second image, the masked regions output on the left is the one I had just opened up, while the one on the right is from my screenshot from earlier. The one on the left is missing several maskings that the one on the right has. They are both the same file, opened just a few minutes apart. Ill do some more testing on this a bit later, but do you have any ideas that may be the reason it is cut short, even after saving? Could there be a conflict between autosaving and regular saving?

It looks like when changed, they change to a different format. Instead of being (number, number), it changes to [number. number.]. The brackets change, and they lose the comma. Sometimes they add a decimal point, but there sometimes aren't numbers after the decimal point, either. Could this cause problems later on? This happens in both head and tail_dorsal.

So, for this issue I masked an area, used the blur, realized I made a mistake, and reset. I masked another area, and changed the blur again, and the old masked area appeared.

It appears to happen if you mask an area, press reset, and then change the blur. If you save after resetting, it still saves the masked area, even without messing with the blur. If you change the blur again, it appears to bring back the masked area into the image.

The only thing different between these two images is I changed the blur

User draws a shape (rectangle, circle, lasso) around edges that they want masked

Changing the points for an individuals tail spine changes the values for their location, but does not update the "tail_spine_length".

I just tested it with a few individuals, but these lengths aren't normal. I didn't have any broken tail individuals in here.

Austin, can we limit the output.txt files to just save the subset that we provide as input? I.e. each output.txt would only save modifications for those 50 pictures that we use as input?

It looks like if you mask an area, then the image needs to be reset, previous maskings/unmaskings are still uploaded to the masked_image_output when you save the final version. Do you think there is anywhere else in the other outputs that may be similar to this, where resetting doesn't fully reset the image?

The saving indicator no longer appears when saving. We think it disappeared around the time of the last git pull or two ago. We are back to the spinny wheel.

Issue with the landmark fit line not showing up when the GUI is first started. Suspect it has to do with path mismatch between the image path and the analysis path.

So, we recreated the crash the other day. We were moving through images in the xag file attached. xag.txt
We reached the file /Users/aob-lab/Documents/Daphnia_phenotyping/pictures/full_101791_D8_137_juju_2B_RigB_20170728T114141.bmp when it crashed. Looking at the output file, it doesn't seem to look different than the others, so we don't quite know why it crashed. Loading it up again (we were able to save right before the image), we got an error that a different file, image /Users/aob-lab/Documents/Daphnia_phenotyping/pictures/full_111705_DBunk_215_juju_1A_RigB_20171130T130516.bmp was not found.

Afterwards, it opened the picture full_100333_AD10_1_juju3_1D_RigB_20170603T123526.bmp
/Users/aob-lab/Documents/Daphnia_phenotyping/pictures/, which is actually further up on the list, and should have already been completed (Since I believe that we move from top to bottom of the file). We don't see it in the output file, so could there be a reason it was skipped originally? When pressing next after that, it moved on to the first file listed here, that ends in T114141, normally (Which had crashed the program). It then went in order fine after that, until it finished.

It looks like we have another crashing issue. We were just paging through the pictures to check the outputs, and when we reached this image, it appears to crash. We have previously found several images that didn't have a shape output, meaning the fit didn't appear, but this one crashes the gui. It also seems like the last checkpoint (Closest to the tail) appears to be fairly resistant to being changed. The input files are on the workstation in /mnt/spicy_4/daphnia/analysis/analysis_results/inputTXTaob-lab, which I had downloaded locally to the Mac and changed to a txt file. This one was in xaa, around the 42nd picture. The outputs we are currently using are different than the ones used previously, as it seems we were missing ~200 images from it. They are no longer sorted from worst-best, but in numerical order.

Quite a number of images that are messed up have something wrong with the automatic masking feature. Would it be too much of an issue to test the gui without it? Even the ones that don't have much of an issue still need to be manually masked/unmasked due to it cutting off odd edges.

When a picture is masked, then the blur is changed, the masking goes away on the image (sort of like it is being reset). The maskings that no longer show up nor influence the checkpoints are still saved in the masked_regions_output file. I'm not sure if this influences the traceback errors Dörthe was getting, but it still is an issue with the blur.

It sounds like it crashed in the middle of using the program when it couldn't find the picture, and we aren't sure if it saved the previous ones. It now will not open up properly. This can probably be resolved by deleting that line, but we would like if there was a "skip" function if it didn't find the picture.

To ID who modified checkpoints

Not sure if this will show up in the excel sheet, but some images may have debris/broken tails. We weren't sure if this may impact it, but we may need a button that can reset the tailtip coordinates to where the tail actually ends.

Accepted images aren't being purged from the clone_list properly. The 'accepted' state and changes seem to be working, but they still show up in the GUI.

I just pulled the newest version (and updated the gui_params.txt), and this is the error I am getting. Any idea, Austin?

Austin, we have something weird happening over here. Robert and I sometimes had the feeling that the GUI is skipping some images. That's why we decided to run every txt input list once again after running through it for modifications. So, we would either have those missing pics popping up or getting the message 'either image list is empty or they have all been 'accepted'.

However, for the xah input file, two of our Macs open 8 pics or so up (1st screenshot), while a third one is giving the 'all accepted' message. We double checked that all drives are properly mounted and that the params.txt file is correct etc. ANY IDEA why this is happening???

I just set skip accepted to 1 for the first time in a while, and it appears to be breaking when you try to skip past images that are missing. This error also occurs when the missing image is first in the block. The issue we had when the last image was missing still appears to be working, though.

Skip accepted seems to work fine when there aren't any missing images that have been saved.

Working through xaa, and have had several images give this error when trying to move checkpoints. It crashes the image itself, so that the picture is no longer shown. I had tried unmasking an area to slightly move one of the checkpoints. One of the images is /full_110258_D8_183_ctrl_2A_RigB_20170817T162711.bmp

Oftentimes the checkpoints don't connect. Suspect it has to do with interpolation (don't even know if it does interpolation currently, might just only fill existing pixel edges)

After encountering the last error (#24), I had saved and exited the program. Now, when I try to resume the session, the program will not start up, and gives this error:

Just an example of stuff that is popping up on terminal while modifying the fits. Most of the time, these warnings and errors still allow modifying and saving properly. However, I thought I would briefly check in whether there is indeed nothing to worry about. THANKS

Terminal Saved Output_xai.txt

can we add a text field to include some additional notes on the pic

Austin, I just briefly wanted to follow up on the validation stuff using the Fiji program. As mentioned before, I would like to only run through all pics once assessing ANIMAL LENGTH and PEDESTAL SCORE at the same time. Could you combine those loops we talked about?

In doing the most recent test, we were going to do the first 20 or so of xaa and xfp to see if it gave any errors, and it didn't. However, it appears that the metadata output file isn't saving the blur, and the masked_regions_output only saved one of the maskings that I did, and nothing else. I believe I saved right after it, so it may be a case of not adding on the additional maskings created after the original save. The outputs should all be marked xaaRPreduced or xfpRPreduced in /mnt/spicy_4/daphnia/analysis/analysis_results. Dörthe is testing this again right now, and Ill also double check it.

DB: in the output file, there are a few columns that turned up anew and/or their input changed (e.g., from numbers to TRUE vector etc); could you have a look at these columns:

tail tip / tail base / anterior_mask_endpoints / deyecenter_pedestal_max / dorsal residual / automated_PF (Robert will include this file in his email).

Before going through the ~7300 pictures, we took a minute to stop and compare the before and after files. Some of the columns do not appear to be getting recalculated after changing and saving the image in the gui (i.e., animal length, etc.). We weren't sure if these should've been changed, as in some of them the fit was pretty badly miscalculated. At least one set of columns (AR and AS when it is changed into an excel sheet) are switched with each other. The tail dorsal coordinates has also gone from a set of coordinates, into a single number. Attached is a single example from one of the images that we had worked on. It had been flipped, and corrected. Could you double check some images to make sure that it is actually recalculating the values that need to be recalculated? The files are daphnia_output_20180926 as the original, and daphnia_output_xaa_20181010 as the output. We have gone all the way to xaf, (xaa, xab...xaf), with 50 images per file. We used the terminal's diff to get several comparisons.

example_pheno-output.txt

DB: can we add a 'save original’ as a 3rd option in addition to ‘accept change’ and ‘save’ in order to differentiate between pics that were changed and those that were not (based on the ‘modified’ column); this is in addition to the previously discussed ‘flag’ button that will exclude any pics that we consider not worth modifying

There seem to be ~320 images that the gui does not like. This was what had caused the crash the other day, not the lack of a stage micrometer. It appears that they seem to be missing quite a bit of information. Could you double check why they are nan's, and why the gui pipeline can't load them in? If thats a huge issue, we can definitely exclude them, but we would like you to check first, because that is over 300 images that we would have to exclude.

@awedwards - can you have a look at the shape_output files, please? I just checked some again, and it seems that there is something wrong. I only realized this because the size of those output files varies quite a bit. And, when looking at the data frame - some files only seem to have data for a few pictures... Any idea why this is? Also, can you recalculate this data from the daphnia_output txt files? (or would we need to re-do our modifications again?)

So we don't have to rely on spinning wheel to know if things are done saving or not

Austin, we set-up another computer, and it seems like the mounted volumes (i.e. that /Volumes/daphnia) is always disappearing, plus I need to do the sshfs command to link it to the mnt/spicy_4 directory. I have no idea why this computer is always unmounting the drive. Do you have any ideas? If there isn't anything that comes up immediately, no worries, I can simply start each round of pictures with some more scripting lines. Just thought to briefly check in with you if there were an easy fix ;-)

Text box for changing blur needs work re: input handling

Each of these could probably be their own issue, but I just wanted a place to have these for when we need to do the second pass of the GUI.

1st: Tail doesn't get snipped properly
Most of the images that need to be tweaked without having anything wrong with the image itself seem like the tail doesn't get snipped properly

2nd: Moving fit.
When you move the checkpoint towards the head, the tail checkpoint moves up the daphnia.

3rd: Fit doubles back on itself.
The last checkpoint sometimes isn't the actual last checkpoint, and the fit doubles back on itself to the second last checkpoint. This is difficult to see, but pretty apparent when we look at the data later. Deleting certain checkpoints fixes this, but its still difficult to see when it happens.

4th: Add checkpoints doesn't work.
We can't use the add checkpoint button. Even when we can add the checkpoint, they disappear when the fit moves next. Checkpoints also have a tendency to disappear randomly.

5th:The fit can't bridge gaps properly.
Sometimes the fit hits a gap, then goes straight to the last checkpoint. This is pretty annoying when the gap is pretty small. We can't add checkpoints to help with this.

Things to add?

1st: Drawing function?
When the edges are on, it would be great if we could "bridge the gap" between two lines. Maybe something that we could manually add edges into the image, and have the pixels we added saved in one of the outputs?
Things to think about: Simply drawing the line may be a bit jagged, but the simplest option. Maybe we could add in something like Powerpoint's drawing tool, where you select 2 points, and it draws a curve between them, which you can then modify the angle of. We'd be able to save a larger number of pictures this way, but I'm not too sure of how feasible it is.

2nd: Absolute areas?
Places where the fit is absolutely not allowed to go? Similar to the masking function, but completely blocks the fit from going there? We would probably need this added into the masking/unmasking file.

I believe the GUI is alright at the moment for what we are doing (we are already a good way through the first passthrough, and I would rather not have to spend time bug testing in the middle of this pass). If we could start thinking about some of these fixes/options for when we are doing the second passthrough, that would be great.

So, on the good side, errors are now saved in the stderr file. However, we cannot see the filename pop up in the terminal anymore. We would prefer having it pop up, just to keep track of where we are...

We are unable to place any checkpoints closer to the eye in some of the images, where the antennae are automatically masked.

Hi Austin,

We are setting up another computer in the office to run the program, and we are now getting this error. Any ideas?

/Users/aob-lab/Library/Python/2.7/lib/python/site-packages/openpyxl/reader/worksheet.py:318: UserWarning: Unknown extension is not supported and will be removed
warn(msg)
Loading 11_12_Inductions_Metadata.xlsx
Loading 12_03_Inductions_Metadata.xlsx
Loading 07_23_Inductions_Metadata.xlsx
Loading 08_13_Inductions_Metadata.xlsx
Loading 07_16_Inductions_Metadata.xlsx
Loading 06_05_Inductions_Metadata.xlsx
Loading 07_09_Inductions_Metadata.xlsx
Loading 06_18_Inductions_Metadata.xlsx
Loading 11_26_Inductions_Metadata.xlsx
Loading 12_10_Inductions_Metadata.xlsx
Loading 05_29_Inductions_Metadata.xlsx
Loading 06_12_Inductions_Metadata.xlsx
Loading 10_23_Inductions_Metadata.xlsx
Loading 07_30_Inductions_Metadata.xlsx
Loading 12_17_Inductions_Metadata.xlsx
Loading 09_18_Inductions_Metadata.xlsx
Loading 10_02_Inductions_Metadata.xlsx
Loading 10_08_Inductions_Metadata.xlsx
Loading 05_22_Inductions_Metadata.xlsx
Loading pond and season metadata

Reading in image file list
Reading in analysis file
Traceback (most recent call last):
File "./daphnia/gui.py", line 760, in
v = Viewer(gui_params, params)
File "./daphnia/gui.py", line 467, in init
clone = utils.dfrow_to_clone(self.data, np.where(self.data.filebase == fileparts[-1])[0][0], self.params)
File "/Users/aob-lab/Library/Python/2.7/lib/python/site-packages/pandas/core/generic.py", line 3081, in getattr
return object.getattribute(self, name)
AttributeError: 'DataFrame' object has no attribute 'filebase'
biol-bergland-IM2011:daphnia aob-lab$

Could make it so that if we press a key (i.e. q) it would be the same as pressing the "Delete checkpoint" button? There are times where we have to delete several checkpoints in a row, and having the tail and tail tip checkpoints makes it a bit more tedious to delete long rows of them. If we could have a key to do the same, that would speed things up quite a bit more.

We started having this issue after the update ~2 days ago.

It looks like the update went well, and it seems to record masking fairly well. It doesn't record any unmasking that we do, though.

Is there any way to differentiate between masking and unmasking, too? I am a bit worried about pictures we need to mask/unmask several times. Maybe is there a way to set an order to the maskings/unmasking (1st masking is 1, 2nd masking is 2, unmasking is 3). That way we can tell when the final masked/unmasked image looked like?

For this picture, I masked the dorsal half of the daphnia, then unmasked the dorsal edge. It only shows the section that was masked.

In this picture, I made a complex shape (a flower) to make sure it was recording it down properly. I then unmasked a section of it (the bar through the center). It recorded the shape well, but not the bar.

Overlapping maskings seems to be fine, with it recording each set properly (S with a bar through it)

Trying to mask pictures that have already been saved gives this error: