Improve show methods #511

rwood-97 · 2024-10-10T09:22:42Z

Summary

Currently we have some very basic "show" methods for veiwing results. This is good for a quick look but would be better if we had some more advanced tools for visualising our results as part of the MapReader library.

Fixes #341
Address #478 - removed the utils.py file from annotate

Describe your changes

Updates the show methods in images.py, runner_base.py and rec_runner_base.py and adds explore_xx methods in each of these classes. Explore uses geopandas explore to create an interactive visualisation of results.
Also adds a check_georeferencing method ot the runners as this is needed to ensure we can convert to geo coords and therefore use the explore methods.
Adds infer_coords_from_patches method in images.py which uses patches to infer parent coords. This is for e.g. if you have loaded patches via the load_patches method and only added patch metadata (e.g. predictions).
Updates the way in which metrics are stored in classifier.py - now instead of train_loss_epoch and train_fscore_epoch as the key for a dict, we have nested dict with keys of train, val, etc and then within that keys of loss, fscore etc.
This enables simplification of the plot_metrics method, users now pass a list of metrics to plot (e.g. "loss", "fscore_micro", etc) and it will plot by default all phases. Can tell it just to plot for one phase by passing the phases argument.
I moved initialise model to the top of the file because it made sense to be grouped with the initialise_optimiser, initialise_scheduler etc methods.

Using the show_patches method:

Using the explore_patches method:

Using explore_predictions method (i.e. the equivalent to explore_patches but for text spotting):

Checklist before assigning a reviewer (update as needed)

Reviewer checklist

Please add anything you want reviewers to specifically focus/comment on.

Everything looks ok?

…emove old show function

…/scheduler/etc, refactoring train_core, update metrics dict and show_metrics functions

codecov · 2024-10-17T14:35:10Z

Codecov Report

Attention: Patch coverage is 71.94093% with 133 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
mapreader/classify/classifier.py	63.19%	60 Missing ⚠️
mapreader/spot_text/runner_base.py	34.88%	28 Missing ⚠️
mapreader/load/images.py	81.25%	24 Missing ⚠️
mapreader/spot_text/rec_runner_base.py	35.71%	18 Missing ⚠️
mapreader/classify/load_annotations.py	0.00%	3 Missing ⚠️

Flag	Coverage Δ
unittests	`77.66% <71.94%> (+3.10%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
mapreader/download/sheet_downloader.py	`83.38% <100.00%> (+0.04%)`	⬆️
mapreader/spot_text/deepsolo_runner.py	`82.14% <ø> (ø)`
mapreader/spot_text/dptext_detr_runner.py	`89.55% <ø> (ø)`
mapreader/spot_text/maptext_runner.py	`82.75% <ø> (ø)`
setup.py	`0.00% <ø> (ø)`
tests/test_classify/test_classifier.py	`100.00% <100.00%> (ø)`
tests/test_load/test_images.py	`100.00% <100.00%> (ø)`
mapreader/classify/load_annotations.py	`59.29% <0.00%> (-0.64%)`	⬇️
mapreader/spot_text/rec_runner_base.py	`73.68% <35.71%> (-2.79%)`	⬇️
mapreader/load/images.py	`88.53% <81.25%> (+9.95%)`	⬆️
... and 2 more

kallewesterling · 2024-10-29T14:19:05Z

I really like this PR, @rwood-97. Haven't had time/been able to test run it yet. Do you need me to do that?

rwood-97 · 2024-11-01T11:56:30Z

@kallewesterling I've updated the docs and two of the worked examples.
Could you potentially run the worked examples to check it all works

Maybe if you have time an ideal thing would be to use the docs to write an example of show_xx and explore_xxx methods (even if just getting up to that point in the worked examples I've updated and then writing your own cells instead of running the ones I've written if that makes sense.

Let me know what you have time for and I could potentially get @thobson88 or @kmcdono2 to check it.

kallewesterling · 2024-11-12T09:56:56Z

@kallewesterling I've updated the docs and two of the worked examples. Could you potentially run the worked examples to check it all works

Maybe if you have time an ideal thing would be to use the docs to write an example of show_xx and explore_xxx methods (even if just getting up to that point in the worked examples I've updated and then writing your own cells instead of running the ones I've written if that makes sense.

Let me know what you have time for and I could potentially get @thobson88 or @kmcdono2 to check it.

Sorry for the late response! I'll check it out and will test run things.

I'll also use the docs to write an example of the show_xx and explore_xxx methods!

kallewesterling · 2024-11-12T10:40:32Z

A few things:

Weirdly, the annotator stopped working. I am running Python 3.9.20

./models_tutorial/checkpoint_10.pkl cannot be found (my last checkpoint is 9):

my_classifier = ClassifierContainer(
    model=None,
    labels_map=None,
    load_path="./models_tutorial/checkpoint_10.pkl",
)

---------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
Cell In[77], [line 1](vscode-notebook-cell:?execution_count=77&line=1)
----> [1](vscode-notebook-cell:?execution_count=77&line=1) my_classifier = ClassifierContainer(
      [2](vscode-notebook-cell:?execution_count=77&line=2)     model=None,
      [3](vscode-notebook-cell:?execution_count=77&line=3)     labels_map=None,
      [4](vscode-notebook-cell:?execution_count=77&line=4)     load_path="./models_tutorial/checkpoint_10.pkl",
      [5](vscode-notebook-cell:?execution_count=77&line=5) )

File ~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:130, in ClassifierContainer.__init__(self, model, labels_map, dataloaders, device, input_size, is_inception, load_path, force_device, **kwargs)
    [125](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:125)     print(
    [126](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:126)         "[WARNING] Ignoring ``labels_map`` as ``load_path`` is specified."
    [127](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:127)     )
    [129](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:129) # load object
--> [130](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:130) self.load(load_path=load_path, force_device=force_device)
    [132](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:132) # add any extra dataloaders
    [133](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:133) if dataloaders:

File ~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1730, in ClassifierContainer.load(self, load_path, force_device)
   [1727](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1727) mydevice = self.device
   [1729](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1729) if not os.path.isfile(load_path):
-> [1730](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1730)     raise FileNotFoundError(f'[ERROR] "{load_path}" cannot be found.')
   [1732](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1732) print(f'[INFO] Loading "{load_path}".')
   [1734](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1734) with open(load_path, "rb") as myfile:
   [1735](https://file+.vscode-resource.vscode-cdn.net/Users/kwesterling/Desktop/temp/november-2024-mapreader/MapReader/worked_examples/geospatial/classification_one_inch_maps/~/Desktop/temp/november-2024-mapreader/MapReader/mapreader/classify/classifier.py:1735)     # objPickle = pickle.load(myfile)

FileNotFoundError: [ERROR] "------------/models_tutorial/checkpoint_10.pkl" cannot be found.

kallewesterling · 2024-11-12T10:45:07Z

Otherwise, looking good!

kallewesterling · 2024-11-12T11:04:47Z

@rwood-97 Where in the docs do you think we should add the parts about these methods? As a new section under the post-processing step?

rwood-97 · 2024-11-12T12:58:04Z

Re. docs, I've added them in the load docs as an extension of the old show methods.

I think the two issues in the worked examples are actually okay, I'm just working on moving these over to the worked-examples repo (dev branch) and have fixed them I thinkl

kallewesterling · 2024-11-12T13:40:23Z

Oh, I see now that you've already added the methods to the documentation. Can you let me know what you need me to do with the docs, then, @rwood-97 ?

rwood-97 · 2024-11-12T13:41:28Z

I think just have a quick look and check it makes sense would be ideal, otherwise I think this is good to merge

kallewesterling · 2024-11-18T17:01:22Z

Wiho!! MERGED!!

rwood-97 added 8 commits October 8, 2024 17:00

remove unused file

30785c7

update plotting in images.py, add infer coords from patches method, r…

030d9e2

…emove old show function

return none if no annotations with specified label found in show_sample

f089310

always add optimiser/scheduler/etc after calling ititialize_optimiser…

fb80834

…/scheduler/etc, refactoring train_core, update metrics dict and show_metrics functions

ensure show works for grayscale images

827cf95

update show and add explore to text spotting runners

8eae31b

allow parent df to be None

8db862b

add check for georeferencing

cee4d21

rwood-97 requested a review from kallewesterling October 10, 2024 09:34

rwood-97 added 2 commits October 10, 2024 10:58

update requirements

456c2ed

fix current tests

21ace16

rwood-97 added 3 commits October 29, 2024 12:01

update tests for infer_coords_from_patches

aa48a19

add tests for classifier

7e5a198

add tests for explore methods

cbf20c2

rwood-97 added 8 commits October 29, 2024 14:45

Merge branch 'main' into improve_show

b9b6f69

add tests georeferencing

40d1da9

Merge branch 'main' into improve_show

a26e87e

update changelog

509edb3

update worked examples and two minor fixes

16f0378

add worked example outputs to gitignore

93ef0a4

update docs

c860b6a

update docs and add list_metrics method to classify

4db7664

rwood-97 added 2 commits November 8, 2024 12:26

minor fix for georeferencing if no parents (i.e. load_patches)

caa4103

Merge branch 'main' into improve_show

b735342

Fixing order of explore_patches if clauses

eede7bb

Merge branch 'main' into improve_show

77187eb

kallewesterling approved these changes Nov 18, 2024

View reviewed changes

rwood-97 merged commit 0999cf3 into main Nov 18, 2024
8 of 9 checks passed

rwood-97 deleted the improve_show branch November 18, 2024 11:54

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve show methods #511

Improve show methods #511

rwood-97 commented Oct 10, 2024 •

edited

Loading

codecov bot commented Oct 17, 2024 •

edited

Loading

kallewesterling commented Oct 29, 2024

rwood-97 commented Nov 1, 2024

kallewesterling commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

rwood-97 commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

rwood-97 commented Nov 12, 2024

kallewesterling commented Nov 18, 2024

Improve show methods #511

Improve show methods #511

Conversation

rwood-97 commented Oct 10, 2024 • edited Loading

Summary

Describe your changes

Checklist before assigning a reviewer (update as needed)

Reviewer checklist

codecov bot commented Oct 17, 2024 • edited Loading

Codecov Report

kallewesterling commented Oct 29, 2024

rwood-97 commented Nov 1, 2024

kallewesterling commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

rwood-97 commented Nov 12, 2024

kallewesterling commented Nov 12, 2024

rwood-97 commented Nov 12, 2024

kallewesterling commented Nov 18, 2024

rwood-97 commented Oct 10, 2024 •

edited

Loading

codecov bot commented Oct 17, 2024 •

edited

Loading