Dataclasses and post-processing refactor #2098

djdameln · 2024-05-31T12:06:12Z

📝 Description

Dataclasses:

Adds new module anomalib.dataclasses, which contains image and batch level dataclasses across different modalities.
The fields of the dataclasses are instances of the new FieldDescriptor class. This allows specifying a validation method for each field that gets called every time the field is updated. This ensures consistency in the typing and shape formatting of the fields of all instances of the dataclass.

Post-processing:

adds new PostProcessor base class and OneClassPostProcessor concrete subclass.
PostProcessor inherits from Callback, so post processing operations can easily be applied in any of the callback hooks.
PostProcessor inherits from nn.Module, so that the post-processing steps can be included in the forward pass of the exported model (onnx/torch/openvino). This means that subclasses of PostProcessor have to implement the forward method. In the exported model, the post processor's forward method is called after the torch model's forward method.
The new OneClassPostProcessor class implements our previous post-processing behaviour, consisting of normalization and thresholding.
Custom post-processing can be achieved by passing a post processor instance to the engine using the post_processor argument.

✨ Changes

Select what type of change your PR is:

🐞 Bug fix (non-breaking change which fixes an issue)
🔨 Refactor (non-breaking change which refactors the code base)
🚀 New feature (non-breaking change which adds functionality)
💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)
📚 Documentation update
🔒 Security update

✅ Checklist

Before you submit your pull request, please make sure you have completed the following steps:

📋 I have summarized my changes in the CHANGELOG and followed the guidelines for my type of change (skip for minor changes, documentation updates, and test enhancements).
📚 I have made the necessary updates to the documentation (if applicable).
🧪 I have written tests that support my changes and prove that my fix is effective or my feature works (if applicable).

For more information about code review checklists, see the Code Review Checklist.

ashwinvaidya17

Let's leave visualizer refactor to a different PR. It might require a lot more changes. We can just add a TODO to the comments in the inferencers.
The rest looks good to me. Once we remove the commented code we can push this as is.

tools/inference/torch_inference.py

review-notebook-app · 2024-09-02T08:48:54Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

codecov · 2024-09-02T10:44:10Z

Codecov Report

Attention: Patch coverage is 90.88472% with 102 lines in your changes missing coverage. Please review.

Please upload report for BASE (feature/design-simplifications@4b1e08c). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/anomalib/dataclasses/generic.py	82.31%	29 Missing ⚠️
src/anomalib/dataclasses/torch.py	90.20%	28 Missing ⚠️
src/anomalib/dataclasses/numpy.py	85.86%	13 Missing ⚠️
src/anomalib/utils/visualization/image.py	84.00%	8 Missing ⚠️
...rc/anomalib/models/video/ai_vad/lightning_model.py	58.33%	5 Missing ⚠️
src/anomalib/models/image/dsr/torch_model.py	66.66%	4 Missing ⚠️
.../anomalib/models/components/base/anomaly_module.py	81.25%	3 Missing ⚠️
src/anomalib/post_processing/one_class.py	96.93%	3 Missing ⚠️
...anomalib/deploy/inferencers/openvino_inferencer.py	91.30%	2 Missing ⚠️
src/anomalib/models/video/ai_vad/torch_model.py	50.00%	2 Missing ⚠️
... and 5 more

Additional details and impacted files

@@                        Coverage Diff                        @@
##             feature/design-simplifications    #2098   +/-   ##
=================================================================
  Coverage                                  ?   80.00%           
=================================================================
  Files                                     ?      250           
  Lines                                     ?    11092           
  Branches                                  ?        0           
=================================================================
  Hits                                      ?     8874           
  Misses                                    ?     2218           
  Partials                                  ?        0

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

* use dataclass for model in- and outputs * split dataclass in image and video * use dataclass in torch inferencer * use dataclass in openvino inferencer * add post_processor class * remove default metrics from CLI * export post processing * add post processor to patchcore * use named tuple for inference outputs * validate and format inputs of PredictBatch * update torch inference * remove base inferencer inheritance * update openvino inference * fix visualization * PredictBatch -> Batch * post processor as callback * use callback methods to apply post processing * temporary fix for visualization * add DatasetItem class * fix pred_score shape and add __len__ * make batch iterable * add in place replace method * use dataset items in inference * dataset_items -> items * use namedtuple as torch model outputs * formatting * split dataclasses into input/output and image/video * merge input and output classes * use init_subclass for attribute checking * add descriptor class for validation * improve error handling * DataClassDescriptor -> FieldDescriptor * add is_optional method * add input validation for torch image and batch * use image and video dataclasses in library * add more validation * add validation * make postprocessor configurable from engine * fix post processing logic * fix data tests * remove detection task type * fix more tests * use separate normalization stats for image and pixel preds * add sensitivity parameters to one class pp * fix utils tests * fix utils tests * remove metric serialization test * remove normalization and thresholding args * set default post processor in base model * remove manual threshold test * fix remaining unit tests * add post_processor to CLI args * remove old post processing callbacks * remove comment * remove references to old normalization and thresholding callbacks * remove reshape in openvino inferencer * export lightning model directly * make collate accessible from dataset * fix tools integration tests * add update method to dataclasses * allow missing pred_score or anomaly_map in post processor * fix exportable centercrop conversion * fix model tests * test all models * fix efficient_ad * post processor as model arg * disable rkde tests * fix winclip export * add copyright notice * add validation for numpy anomaly map * fix getting started notebook * remove hardcoded path * update dataset notebooks * update model notebooks * fix logging notebooks * fix model notebook

djdameln added 2 commits May 30, 2024 17:08

use dataclass for model in- and outputs

68c5582

split dataclass in image and video

ddfcd5f

djdameln requested review from samet-akcay and ashwinvaidya17 as code owners May 31, 2024 12:06

samet-akcay marked this pull request as draft May 31, 2024 12:09

djdameln added 2 commits July 4, 2024 00:02

use dataclass in torch inferencer

32e038d

use dataclass in openvino inferencer

675dd3f

ashwinvaidya17 reviewed Jul 10, 2024

View reviewed changes

tools/inference/torch_inference.py Outdated Show resolved Hide resolved

djdameln added 21 commits July 22, 2024 19:06

add post_processor class

5779ab7

remove default metrics from CLI

0662558

export post processing

fddbeb1

add post processor to patchcore

e32bd7d

use named tuple for inference outputs

90265e8

validate and format inputs of PredictBatch

e3a9c1d

update torch inference

89f972c

remove base inferencer inheritance

08bdae2

update openvino inference

2bc76fc

fix visualization

f7c7f9a

PredictBatch -> Batch

4160ab3

post processor as callback

fd9eb24

use callback methods to apply post processing

87facb6

temporary fix for visualization

2269a78

add DatasetItem class

9652b9f

fix pred_score shape and add __len__

082bbbc

make batch iterable

dbabb20

add in place replace method

b190cd3

use dataset items in inference

ed904eb

dataset_items -> items

773e54a

use namedtuple as torch model outputs

f8d999a

ashwinvaidya17 self-requested a review July 31, 2024 11:48

remove reshape in openvino inferencer

039be2a

samet-akcay marked this pull request as ready for review August 29, 2024 11:13

djdameln added 5 commits August 29, 2024 13:48

export lightning model directly

381e638

make collate accessible from dataset

daead5b

fix tools integration tests

987abe5

add update method to dataclasses

a709c6c

allow missing pred_score or anomaly_map in post processor

a37fa3b

samet-akcay changed the base branch from main to feature/design-simplifications August 29, 2024 19:26

djdameln added 13 commits August 30, 2024 13:27

fix exportable centercrop conversion

14da4fa

fix model tests

beb3b97

test all models

a9d07db

fix efficient_ad

6bcca36

post processor as model arg

014cb59

disable rkde tests

58df063

fix winclip export

25845fb

add copyright notice

1defdba

add validation for numpy anomaly map

8d60276

fix getting started notebook

0afb6d9

remove hardcoded path

a26efb9

update dataset notebooks

e7d9852

update model notebooks

a4bcbfe

djdameln added 3 commits September 2, 2024 11:50

Merge branch 'feature/design-simplifications' into refactor_outputs

085c4aa

fix logging notebooks

eff1f97

fix model notebook

40bb4be

samet-akcay approved these changes Sep 2, 2024

View reviewed changes

samet-akcay merged commit a298a3a into openvinotoolkit:feature/design-simplifications Sep 2, 2024
7 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Dataclasses and post-processing refactor #2098

Dataclasses and post-processing refactor #2098

djdameln commented May 31, 2024 •

edited

Loading

ashwinvaidya17 left a comment

review-notebook-app bot commented Sep 2, 2024

codecov bot commented Sep 2, 2024

Dataclasses and post-processing refactor #2098

Dataclasses and post-processing refactor #2098

Conversation

djdameln commented May 31, 2024 • edited Loading

📝 Description

✨ Changes

✅ Checklist

ashwinvaidya17 left a comment

Choose a reason for hiding this comment

review-notebook-app bot commented Sep 2, 2024

codecov bot commented Sep 2, 2024

Codecov Report

djdameln commented May 31, 2024 •

edited

Loading