Start release branch for OpenVINO 2021.3 (openvinotoolkit#2169)

* Update demos/python_demos/formula_recognition_demo/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* upd readme

* use named constant for resolution setter

* Update demos/python_demos/formula_recognition_demo/formula_recognition_demo.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* move formula_rec_demo according to new layout

* regnetx-3.2gf: use the weights file that we download

Previously, it wasn't used in any way; the model constructor just downloaded
the weights anew.

* fastseg-*: use our ONNX converter instead of the one from the repo

The only differences in how they do the conversion compared to us is:

* they use the keep_initializers_as_inputs option, which doesn't do anything
  helpful (it just adds useless inputs to the ONNX model, which are stripped
  by MO);

* they use the verbose option, which dumps a lot of unnecessary information
  to stdout.

Using our own script is:

* shorter in terms of lines of code;
* more consistent with other PyTorch models;
* more flexible in that we could alter the output by adjusting the ONNX
  converter (e.g. by changing to a newer opset).

This change has no effect on the resulting IRs.

* Add the output recording to the file

* Mzuevx/rnnt evaluator (openvinotoolkit#1993)

* Introduce encoder-prediction-joint evaluator

* speech_recognition_demo: remove unnecessary code from ctcdecode_numpy

This has two main parts:

* Removing unnecessary includes and directives from decoders.i, so that we
  don't generate bindings that are never used.

* Removing the numpy_beam_decode_no_lm function, which is useless: it's
  equivalent to numpy_beam_decode with scorer = nullptr, so in Python one
  can just call numpy_beam_decode with scorer=None.

Altogether, this removes over 3000 lines of code.

* Update smart_classroom_demo

* remove unnecessary casts

* restore cast to float in filterBBoxes due to warning

* Modify C++ images_capture

* AC: simplify CTCGreedyDecoder._ctc_decoder_prediction() (openvinotoolkit#2044)

* formula_recognition_demo: remove extra directory

Looks like an artifact of the demo layout change. The README expects
these files to be in the main demo directory.

* Update instance segmentation (openvinotoolkit#1719)

* AC: support validation scheme API (openvinotoolkit#2041)

* AC: support validation scheme API

* add tests for API

* bound coonfig validation and scheme

* bump version

* AC: support YOLO V3 tf2 (openvinotoolkit#2046)

* ci/update-requirements.py: update to reflect recent project changes

* The demo requirements file is now in a different location.

* We need to do additional processing to remove the OpenVINO install path,
  because pip-compile writes the pot dependency as "pot (/opt/intel/openvino...)".

* Bump all CI requirement versions

The notable changes are:

* Updating PyYAML to 5.4.1, which fixes a long-standing security vulnerability
  (which we're not affected by, but still).
* Updating NumPy to 1.19.5, which works around the Windows bug that previously
  forced us to downgrade it to 1.19.3.

* ci/requirements-*: regenerate with new pip-tools

There's no change in versions, but the formatting is different.

* Apply fixes from review

* AC: supports DETR (openvinotoolkit#2049)

* add ssh-mxnet support

* upd nms and models list

* Added documentation deps generation.

* Added requirements-documentation.txt file.

* Update exception

* add yolov5m config (openvinotoolkit#2051)

* upd comment in main.cpp

* add ssh-mxnet to test cases

* small fix in parameter passing in nms

* Integrated documentation deps to check-basics.

* Delete requirements-documentation.txt

* Update requirements-check-basics.txt

* Deleted trailing whitespace.

* AC: fix config validation for dict field (openvinotoolkit#2053)

* AC: support subset dumping/selection using file (openvinotoolkit#2056)

* AC: support subset dumping/selection using file

* AC: backward compatibility and  fix for data provider size (openvinotoolkit#2061)

* add iou_t to show usage, add comment for nms func

* Update demos/common/cpp/models/src/detection_model_retinaface.cpp

Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

* AC: fix dumping prediction to annotation (openvinotoolkit#2062)

* AC: fix dumping prediction to annotation

* Update quantization_model_evaluator.py

* AC: fix detection adapters which does not produce labels (openvinotoolkit#2064)

* AC: fix annotation reloading (openvinotoolkit#2065)

* AC: fix annotation reloading

* Update tools/accuracy_checker/accuracy_checker/evaluators/quantization_model_evaluator.py

Co-authored-by: Anna Mironova <anna.mironova@intel.com>

Co-authored-by: Anna Mironova <anna.mironova@intel.com>

* AC: fix getting labels (openvinotoolkit#2066)

* add aclnet-int8 model

* enable model in demo

* AC: allow setting device config if device unknown (openvinotoolkit#2067)

* AC: allow setting device config if device unknown

* update warning message

* security_barrier_camera: downgrade frameCounter to 32-bit

Close openvinotoolkit#2040

* Removed unused dependency

* model source link added

* Update demos/text_to_speech_demo/python/utils/text_preprocessing.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* AC: fix quantization for RNNT (openvinotoolkit#2073)

* AC: fix quantization for RNNT

* text-spotting-0004

* Update cases.py

* Update index.md

* Added NetVLAD model (openvinotoolkit#2021)

* Added NetVLAD model

* Aa/melgan forward tacotron att (openvinotoolkit#2024)

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* Update models/intel/text-spotting-0004/description/text-spotting-0004.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* add gflops/mparams

* rm not relevant change

* AC: fix single image processing (openvinotoolkit#2080)

* AC: fix identifier for bert emb (openvinotoolkit#2079)

* AC: fix im2latex evaluator (openvinotoolkit#2081)

* AC: fix im2latex evaluator

* Update im2latex_evaluator.py

* Update im2latex_evaluator.py

* change converter output folder for model quantized in source framework

* AC: Added groups of arguments for AC (openvinotoolkit#2077)

* Added groups of arguments for AC

* Fix comments

* Small fixes

* Small fixes in place_recognition_demo

* AC: update input configuration validation (openvinotoolkit#2083)

* AC: minor issues fixes (openvinotoolkit#2084)

* AC: minor issues fixes

* Apply suggestions from review

* AC: update cocosnet evaluator for quantization (openvinotoolkit#2086)

* AC: update cocosnet evaluator for quantization

* AC: adopt hit ratio representation to work with new identifier (openvinotoolkit#2087)

* Move videowriter

* ci/prepare-documentation.py: add xml:id attributes to top-level sections

* AC: Added field for normalization arguments (openvinotoolkit#2085)

* Added field for normalization arguments

* Fix pylint

* Fix WMT converter (openvinotoolkit#2091)

* Added demo code for ForwardTacotron+Melgan

* Fixed bug in MelGAN inference

* Added extra sequence len for improving quality of generated audio

* Fixed bug

* Fixed splitting of input sequence

* Added support of ForwardTacotron with MultyHeadAttention

* Added defaut width for melgan network

* Updated README.md

* Fixed PEP warnings

* removed unnecessary __init__

Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

* Splitted the logick of decoders

* Changed argparse parameters checking

* Removed redundant backslash

* 1) Changed main return code.
2) Updated README.

* add i3d-rgb model support to action recoginition demo

* Fixed number of arguments for predict method (openvinotoolkit#2092)

* add model to test cases

* add text recognition 0013 model (openvinotoolkit#2078)

* fix code style

* deleted trailing whitespaces

* change model.precision

* make pylint happy

* Update tools/downloader/converter.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* ci/prepare-documentation.py: remove the hardcoded task type list

I originally added the list so that if a model with a new task type
is added, the prepare-documentation script crashes and so the contributor
has to consider what the human-readable description of the task type should
be.

However, given that:

1. for almost all task types, the default string transformation yields a good
   description, and

2. the cost of getting the description wrong is just an awkwardly-named section
   in the documentation TOC,

I don't think that the cost of keeping an extra copy of the task type list
is justified. Instead we should just instruct contributors to name task types
so that the default transformation gives a good result, or, if that's not
possible, to add an override.

* human_pose_estimation/py: change output_limit checking

* ModelFaceBoxes: add a missing variable initializer

* change arg parameter name

* delete whitespaces

* move checks to deserialize

* rm unnecessary import

* fix typo in test cases

* AC: support attention ocr model (openvinotoolkit#2102)

* add nmt models for en-de pair (openvinotoolkit#2097)

* Ik/text spotting 0004 ac (openvinotoolkit#2098)

* change retina to retinaface in args in python demo

* Update tools/downloader/common.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update tools/downloader/common.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* fix unused params, move aclnet classes.txt

* Update models/public/aclnet-int8/aclnet-int8.md

Co-authored-by: Ekaterina Aidova <ekaterina.aidova@intel.com>

* renamed aclnet classes file

* small fixes

* AC: add possibility to add out blob for DNA model (openvinotoolkit#2104)

* add class id to model description (openvinotoolkit#2101)

* update aclnet for OpenVINO 2021.3 release

* update input layer name

* update input name

* add gflops/mparams

* fix links to classes and polished some phrases

* Speech recognition demo: Refactor yoklm lib for n-gram LM inference in ctcdecode_numpy package

* Human pose estimation models update (openvinotoolkit#2063)

* Add DeblurGAN-v2 model (openvinotoolkit#1911)

* Custom evaluator with data feedback (openvinotoolkit#2099)

* update LPR model input according spec (openvinotoolkit#2109)

* Support for YOLO v4 model in py/object_detection_demo

* Fix

* Update models.lst

* Fix

* one more update (openvinotoolkit#2106)

* apply review suggestions

* Added support for using MSCOCO dataset with VOC labels (openvinotoolkit#2105)

* Add new person-vehicle-bike detection models  (openvinotoolkit#2082)

* add draft readme for PVB models

* replace brackets with ``

* add accuracy_checker configs

* add TBD values

* fix accuracy drop

* add unit tests for models

* remove extra spaces

* remove device

* delete cast to np.float32 (openvinotoolkit#2112)

* AC: restore adapter for old text spotting models (openvinotoolkit#2111)

* AC requirements: remove parasail version restriction (openvinotoolkit#2114)

* Fix computation of audio features in AC for mozilla-deepspeech-* that was broken by 28e5a72 (openvinotoolkit#2117)

* Human Pose Estimation C++ demo fixes and update (openvinotoolkit#2110)

* AC def: fix indention in machine translation datasets (openvinotoolkit#2118)

* Address reviewer's comments and fix a bug

1. Fix type of index_limit_ (which should have caused an exception for very large kenlm files).
2. Improve range checking in MemorySectionBitArray to exclude possibility of SIGSEGV in a very exotic crafted case.
3. Add some comments explaining actual ranges, expected ranges and range checking.
4. Fail earlier on strange kenlm files with zero words (1-grams) in vocabulary.

* optional casting to float in grayscale (openvinotoolkit#2122)

* optional casting to grayscale

* pylint warnings

* add parameters to greedy decoder to make it able run tr-0013 (openvinotoolkit#2124)

* speech_recognition_demo: avoid yaml.load

Use safe_load, which can't construct arbitrary Python objects.

* demo updates

* update info about text-rec-13 since model weights and arch changed (openvinotoolkit#2128)

* Apply suggestions from code review

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* very last fix for text-spotting-0004 (openvinotoolkit#2120)

* fix

* added postprocessor

* revert some changes back

* Update text_spotting_evaluator.py

* changed to en-mean, add conversion to RGB

* AC: Fix temporary file usage in MSCOCOorigBaseMetric (openvinotoolkit#2129)

* create the files securely, by using NamedTemporaryFile (previously,
  the file created by NamedTemporaryFile was unused).

* delete the files after they are no longer needed.

* formula_recognition_demo: fix temporary file handling

* create the temporary file securely by using NamedTemporaryFile
  (previously, the file created with NamedTemporaryFile was unused).

* delete the file after use.

* demos: make sure input data lives during inference

Only human_pose_estimation crashed because it pads an image which
creates new Mat and other demos keep unchanged data in `ImageMetaData`.

* downloader: make sure to delete staging files in the download cache

* human_pose_estimation: warn about chosen model size and image size mismatch

* micro changes for deblurring demo

* onwn -> own

* padding and resize

* Update demos/common/python/models/yolo.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update demos/common/python/models/yolo.py

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* vloginov/update text rec 0013 (openvinotoolkit#2130)

* update info about text-rec-13 since model weights and arch changed

* delete targets output

* fix non markdown links

* Update demos/gesture_recognition_demo/python/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update demos/gesture_recognition_demo/python/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update demos/text_to_speech_demo/python/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update demos/text_to_speech_demo/python/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update models/public/aclnet-int8/aclnet-int8.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* intendation

* wrap file refs with dittos

* Update yolo.py

* AC: fix typos in argument names

* tools/downloader: print signal name when a child processes is terminated by one

If a child process is terminated by a signal, there is usually no indication
that that happened; it just looks like it exited normally. This is
confusing for users looking at the conversion log.

Examine the process exit status and, if it was terminated by a signal,
print the signal name and number.

* Added age_recognition adapter (openvinotoolkit#2137)

* Fix spelling errors

* AC: Fix the spelling of "deterministic" (openvinotoolkit#2144)

* object_detection: remove put_highlighted_text()

* Fix yolo TF issue (openvinotoolkit#2147)

* cast float to int for cv::circle arg (openvinotoolkit#2152)

* provide coco labels for different model variants (openvinotoolkit#2146)

* fix CVS-16351 (openvinotoolkit#2141)

* smart_classroom: allow running without dynamic batch (openvinotoolkit#2157)

This PR add safeguard check if model can be processed in dynamic batch.
According to OpenVINO
[documentation](https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_DynamicBatching.html)
dynamic batch feature not supported for models which have:

> layers that might arbitrary change tensor shape (such as Flatten,
Permute, Reshape), layers specific to object detection topologies
(ROIPooling, ProirBox, DetectionOutput), and custom layers

* AC tests: skip affinity setting test (openvinotoolkit#2158)

* object_detection_demo.py: add ctpn wrapper (openvinotoolkit#2076)

* Add ctpn wrapper

* Fix preprocessing for high resolution images

* Rename functions

* Revert changes in main.py, add --input_size key

* Small fixes

* Add condition to consider default input size

* Apply fixes from review

* Modify the code

* Rebase update

* Modify the code of CTPN adapter in AC

* Delete whitespaces

* Fix warning by AC

* add pointers to demo location at demos README.md (openvinotoolkit#2119)

* add pointers to demo location at demos README.md

* Update demos/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* Update demos/README.md

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* rephrase

* improve phrasing

Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>

* demos: fix instance segmentation demo for yolact (openvinotoolkit#2162)

* AC: Added adapter for Ultra-Lightweight Face Detection models (openvinotoolkit#2155)

* AC fix batch in model (openvinotoolkit#2164)

* AC: fix text-to-speech quantization pipeline (openvinotoolkit#2165)

* simple mtcnn demo (openvinotoolkit#2138)

* Face recognition python demo using mtcnn and facenet

Implement the most typical facenet sample.
--
Re-submit to solve the problem below:

you have non relevant commits, because it seems demo developed on top of some branch you worked on for human pose estimation demo

source branch you are based on is quite old (almost a year old), you need to rebase to the latest develop branch

there is an issue if one will try to download models with using models.lst file, at least in case of mtcnn models

you bring some pictures, so need to specify an original source and license for this content

--
If there are still other problems, please remind me.

* Add files via upload

* Add files via upload

* Update README.md

* rebased to the latest develop, remove face reid and images with unclear license

* add to index and test cases

* add sample demo screenshot

* fix pre-commit checks

* fix other pre-commit checks

* fix yet another pre-commit checks

* Update demos/tests/cases.py

* Update demos/tests/cases.py

* Update demos/face_detection_mtcnn_demo/python/face_detection_mtcnn_demo.py

Co-authored-by: Eduard Zamaliev <eduard.zamaliev@intel.com>

* Update demos/face_detection_mtcnn_demo/python/face_detection_mtcnn_demo.py

Co-authored-by: Eduard Zamaliev <eduard.zamaliev@intel.com>

* Update demos/face_detection_mtcnn_demo/python/face_detection_mtcnn_demo.py

Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

* Update demos/face_detection_mtcnn_demo/python/face_detection_mtcnn_demo.py

Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

* modifications according review comments

* remove unused import os

* Fix

* Make executable

* Refactoring

* Fix

Co-authored-by: playboy007gy <50794318+playboy007gy@users.noreply.github.com>
Co-authored-by: Eduard Zamaliev <eduard.zamaliev@intel.com>
Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

* AC: fix single human pose estimation (openvinotoolkit#2167)

* AC: fix single human pose estimation

* Update __init__.py

* Update tools/accuracy_checker/accuracy_checker/data_readers/__init__.py

* AC: fix SQUAD dataset conversion for multi span examples (openvinotoolkit#2168)

* Maskai processing implementation (openvinotoolkit#2160)

* OpenVINO 2021.3 preview candidate models (openvinotoolkit#2166)

* add links to 2021.3 models

* remove EOL models

* remove old ac cfg links

* fix test cases

* Update demos/tests/cases.py

Co-authored-by: Eduard Zamaliev <eduard.zamaliev@intel.com>

* Update demos/tests/cases.py

* Fix documentation for text_detection_demo

* Fix

* Update demos/tests/cases.py

Co-authored-by: Eduard Zamaliev <eduard.zamaliev@intel.com>

* Add CPU performance tuning 'KEY_CPU_THROUGHPUT_STREAMS' to multi_chan… (openvinotoolkit#2153)

* Add CPU performance tuning 'KEY_CPU_THROUGHPUT_STREAMS' to multi_channel demo. Because default stream is 1, which is latency-oriented, it's not optimized for multi_channel situation, which require the throughput-oriented situation.

* Update demos/multi_channel_common/cpp/graph.cpp

Update according to comments

Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

* Update demos/multi_channel_common/cpp/graph.cpp

Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>

Co-authored-by: Vladimir Loginov <vladimir.loginov@intel.com>
Co-authored-by: Roman Donchenko <roman.donchenko@intel.com>
Co-authored-by: Pankratova <anzhella.pankratova@intel.com>
Co-authored-by: mzuevx <60778210+mzuevx@users.noreply.github.com>
Co-authored-by: ivanvikhrev <l4ind@mail.ru>
Co-authored-by: Zlobin Vladimir <vladimir.zlobin@intel.com>
Co-authored-by: Dmitry Sidnev <dmitry.sidnev@intel.com>
Co-authored-by: Nikita Ignatovsky <73649020+nignatovsky@users.noreply.github.com>
Co-authored-by: Aleksei Korobeinikov <aleksei.korobeinikov@intel.com>
Co-authored-by: ivikrhev <ivan.vikhrev@intel.com>
Co-authored-by: Anna Mironova <anna.mironova@intel.com>
Co-authored-by: Vladimir Dudnik <vladimir.dudnik@intel.com>
Co-authored-by: Vladimir <10669582+Wovchena@users.noreply.github.com>
Co-authored-by: andreyan <andrey.anufriev@intel.com>
Co-authored-by: a-a-egorovich <a.a.egorovich@gmail.com>
Co-authored-by: Ilya Krylov <ilya.krylov@intel.com>
Co-authored-by: Ilya Krylov <krylov.ib@gmail.com>
Co-authored-by: Sergei Belousov <belbes122@yandex.ru>
Co-authored-by: Alexey Kruglov <Alexey.Kruglov@intel.com>
Co-authored-by: Pavel Druzhkov <druzhkov.paul@gmail.com>
Co-authored-by: eizamaliev <eduard.zamaliev@intel.com>
Co-authored-by: Galina Zalesskaya <galina.zalesskaya@intel.com>
Co-authored-by: Alexey Kruglov <Alexey.Kruglov@gmail.com>
Co-authored-by: playboy007gy <50794318+playboy007gy@users.noreply.github.com>
Co-authored-by: Yufei Wu <yufei.wu@intel.com>

Author: 23 people

.flake8

@@ -0,0 +1,39 @@
+[flake8]
+exclude=
+    # generic excludes
+    .git,__pycache__,
+    # generated files
+    demos/speech_recognition_demo/python/ctcdecode-numpy/ctcdecode_numpy/impl.py,
+    models/public/mozilla-deepspeech-0.8.2/mds_convert_utils/memmapped_file_system_pb2.py,
+
+select=
+    # indentation problems
+    E10,E11,
+    # closing bracket does not match indentation of opening brackets line
+    E123,
+    # whitespace before '('
+    E211,
+    # missing whitespace after ',', ';', or ':'
+    E231,
+    # missing whitespace after keyword
+    E275,
+    # expected 1 blank line, found 0
+    E301,
+    # blank lines found after function decorator
+    E304,
+    # the backslash is redundant between brackets
+    E502,
+    # statement style issues
+    E7,
+    # parse errors
+    E9,
+    # pyflakes checks
+    F4,F5,F6,F7,F8,F9,
+    # deprecation warnings
+    W6,
+
+ignore=
+    # multiple statements on one line (colon/def)
+    # ignored because putting a statement after a colon is harmless and
+    # sometimes results in more compact code
+    E701,E704,

.gitattributes

@@ -7,8 +7,9 @@
 # so allow it.
 *.patch whitespace=-blank-at-eol
 
-.git* -omz.package
 .editorconfig -omz.package
+.flake8 -omz.package
+.git* -omz.package
 .yamllint -omz.package
 /ci/** -omz.package
 /demos/build_demos.sh omz.package=l,m
@@ -21,6 +22,7 @@
 /ci/requirements-conversion.txt omz.ci.job-for-change.demos omz.ci.job-for-change.models
 /ci/requirements-demos.txt omz.ci.job-for-change.demos
 /ci/requirements-downloader.txt omz.ci.job-for-change.demos omz.ci.job-for-change.models
+/ci/requirements-quantization.txt omz.ci.job-for-change.downloader
 
 /demos/** omz.ci.job-for-change.demos
 /demos/**/*.md -omz.ci.job-for-change.demos