DL models as serverless functions (#1767)

* Initial experiments with nuclio

* Update nuclio prototype

* Improve nuclio prototype for dextr.

* Dummy lambda manager

* OpenFaaS prototype (dextr.bin and dextr.xml are empty).

* Moved openfaas prototype.

* Add comments

* Add serializers and HLD for lambda_manager

* Initial version of Mask RCNN (without debugging)

* Initial version for faster_rcnn_inception_v2_coco

* Fix faster_rcnn_inception_v2_coco

* Implemented mask_rcnn_inception_resnet_v2_atrous_coco

* Implemented yolo detector as a lambda function

* Removed dextr app.

* Added types for each function (detector and interactor)

* Initial version of lambda_manager.

* Implement a couple of methods for lambda:

GET /api/v1/lambda/functions
GET /api/v1/lambda/functions/public.dextr

* First working version of dextr serverless function

* First version of dextr which works in UI.

* Modify omz.public.faster_rcnn_inception_v2_coco

- image decoding
- restart policy always for the function

* Improve omz.public.mask_rcnn_inception_resnet_v2_atrous_coco

* Improve omz.public.yolo-v3-tf function

* Implemented the initial version of requests for lambda manager.

* First working version of POST /api/v1/lambda/requests

* Updated specification of function.yaml (added labels and used annotations section).

* Added health check for containers (nuclio dashboard feature)

* Read labels spec from function.yaml.

* Added settings for NUCLIO

* Fixed a couple of typos. Now it works in most cases.

* Remove Plugin REST API

* Remove tf_annotation app (it will be replaced by serverless function)

* Remove tf_annotation and cuda components

* Cleanup docs and Dockerfile from CUDA component.

* Just renamed directories inside serverless

* Remove redundant files and code

* Remove redundant files.

* Remove outdated files

* Remove outdated code

* Delete reid app and add draft of serverless function for reid.

* Model list in UI.

* Fixed the framework name (got it from lambda function).

* Add maxRequestBodySize for functions, remove redundant code from UI for auto_annotation.

* Update view of models page.

* Unblock mapping for "primary" models.

* Implement cleanup flag for lambda/requests and labeling mapping for functions.

* Implement protection from running multiple jobs for the same task.

* Fix invocation of functions in docker container.

* Fix Dockerfile.ci

* Remove unused files from lambda_manager

* Fix codacy warnings

* Fix codacy issues.

* Fix codacy warnings

* Implement progress and cancel (aka delete) operation.

* Send annotations in batch.

* Fix UI. Now it can retrieve information about inference requests in progress.

* Update CHANGELOG.md

* Update cvat-ui version.

* Update nuclio version.

* Implement serverless/tensorflow/faster_rcnn_inception_v2_coco

* Add information how to install nuclio platform and run serverless functions.

* Add installation instructions for serverless functions.

* Update OpenVINO files which are responsible for loading network

* relocated functions

* Update dextr function.

* Update faster_rcnn function from omz

* Fix OpenVINO Mask-RCNN

* Fix YOLO v3 serverless function.

* Dummy serverless functions for a couple of more OpenVINO models.

* Protected lambda manager views by correct permissions.

* Fix name of Faster RCNN from Tensorflow.

* Implement Mask RCNN via Tensorflow serverless function.

* Minor client changes (#1847)

* Minor client changes

* Removed extra code

* Add reid serverless function (no support in lambda manager).

* Fix contribution guide.

* Fix person-reidentification-retail-300 and implement text-detection-0004

* Add semantic-segmentation-adas-0001

* Moving model management to cvat-core (#1905)

* Squached changes

* Removed extra line

* Remove duplicated files for OpenVINO serverless functions.

* Updated CHANGELOG.md

* Remove outdated code.

* Running dextr via lambda manager (#1912)

* Deleted outdated migration.

* Add name for DEXTR function.

* Fix restart policy for serverless functions.

* Fix openvino serverless functions for images with alpha channel

* Add more tensorflow serverless functions into deploy.sh

* Use ID instead of name for DEXTR (#1926)

* Update DEXTR function

* Added source "auto" inside lambda manager for automatic annotation.

* Customize payload (depends on type of lambda function).

* First working version of REID (Server only).

* Fix codacy warnings

* Avoid exception during migration (workaround)

File "/usr/local/lib/python3.5/dist-packages/django/db/utils.py", line 89, in __exit__
    raise dj_exc_value.with_traceback(traceback) from exc_value
  File "/usr/local/lib/python3.5/dist-packages/django/db/backends/utils.py", line 84, in _execute
    return self.cursor.execute(sql, params)
django.db.utils.ProgrammingError: table "engine_pluginoption" does not exist

* Add siammask serverless function (it doesn't work, need to serialize state)

* Run ReID from UI (#1949)

* Removed reid route in installation.md

* Fix a command to get lena image in CONTRIBUTION guide.

* Fix typo and crash in case a polygon is a line.

Co-authored-by: Boris Sekachev <[email protected]>

.gitattributes

@@ -15,7 +15,6 @@ LICENSE         text
 *.conf          text
 *.mimetypes     text
 *.sh            text eol=lf
-components/openvino/eula.cfg           text eol=lf
 
 *.avi           binary
 *.bmp           binary

.github/CODEOWNERS

@@ -23,7 +23,7 @@
 /datumaro/      @zhiltsov-max
 /cvat/apps/dataset_manager/ @zhiltsov-max
 
-# Advanced components (e.g. OpenVINO)
+# Advanced components (e.g. analytics)
 /components/    @azhavoro
 
 # Infrastructure

.gitignore

@@ -7,7 +7,6 @@
 /.env
 /keys
 /logs
-/components/openvino/*.tgz
 /profiles
 /ssh/*
 !/ssh/README.md

.pylintrc

@@ -20,7 +20,7 @@ persistent=yes
 
 # List of plugins (as comma separated values of python modules names) to load,
 # usually to register additional checkers.
-load-plugins=
+load-plugins=pylint_django
 
 # Use multiple processes to speed up Pylint.
 jobs=1
@@ -66,8 +66,8 @@ enable= E0001,E0100,E0101,E0102,E0103,E0104,E0105,E0106,E0107,E0110,
         W0122,W0124,W0150,W0199,W0221,W0222,W0233,W0404,W0410,W0601,
         W0602,W0604,W0611,W0612,W0622,W0623,W0702,W0705,W0711,W1300,
         W1301,W1302,W1303,,W1305,W1306,W1307
-        R0102,R0201,R0202,R0203
-        
+        R0102,R0202,R0203
+
 
 # Disable the message, report, category or checker with the given id(s). You
 # can either give multiple identifiers separated by comma (,) or put this

.vscode/launch.json

@@ -27,7 +27,7 @@
             "request": "launch",
             "stopOnEntry": false,
             "justMyCode": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath": "${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "runserver",
@@ -58,7 +58,7 @@
             "request": "launch",
             "stopOnEntry": false,
             "justMyCode": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath": "${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "rqworker",
@@ -77,7 +77,7 @@
             "request": "launch",
             "stopOnEntry": false,
             "justMyCode": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath": "${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "rqscheduler",
@@ -93,7 +93,7 @@
             "request": "launch",
             "justMyCode": false,
             "stopOnEntry": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath":"${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "rqworker",
@@ -112,7 +112,7 @@
             "request": "launch",
             "justMyCode": false,
             "stopOnEntry": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath": "${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "update_git_states"
@@ -128,7 +128,7 @@
             "request": "launch",
             "justMyCode": false,
             "stopOnEntry": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath": "${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "migrate"
@@ -144,7 +144,7 @@
             "request": "launch",
             "justMyCode": false,
             "stopOnEntry": false,
-            "pythonPath": "${config:python.pythonPath}",
+            "pythonPath": "${command:python.interpreterPath}",
             "program": "${workspaceRoot}/manage.py",
             "args": [
                 "test",

CHANGELOG.md

@@ -6,6 +6,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [1.1.0-beta] - Unreleased
 ### Added
+- DL models as serverless functions (<https://github.com/opencv/cvat/pull/1767>)
 - Source type support for tags, shapes and tracks (<https://github.com/opencv/cvat/pull/1192>)
 - Source type support for CVAT Dumper/Loader (<https://github.com/opencv/cvat/pull/1192>)
 - Intelligent polygon editing (<https://github.com/opencv/cvat/pull/1921>)
@@ -14,13 +15,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 - Smaller object details (<https://github.com/opencv/cvat/pull/1877>)
+- It is impossible to submit a DL model in OpenVINO format using UI. Now you can deploy new models on the server using serverless functions (<https://github.com/opencv/cvat/pull/1767>)
 - Files and folders under share path are now alphabetically sorted
 
 ### Deprecated
 -
 
 ### Removed
--
+- Removed OpenVINO and CUDA components because they are not necessary anymore (<https://github.com/opencv/cvat/pull/1767>)
 
 ### Fixed
 - Some objects aren't shown on canvas sometimes. For example after propagation on of objects is invisible (<https://github.com/opencv/cvat/pull/1834>)
@@ -82,12 +84,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Colorized object items in the side panel (<https://github.com/opencv/cvat/pull/1753>)
 - [Datumaro] Annotation-less files are not generated anymore in COCO format, unless tasks explicitly requested (<https://github.com/opencv/cvat/pull/1799>)
 
-### Deprecated
--
-
-### Removed
--
-
 ### Fixed
 - Problem with exported frame stepped image task (<https://github.com/opencv/cvat/issues/1613>)
 - Fixed dataset filter item representation for imageless dataset items (<https://github.com/opencv/cvat/pull/1593>)