Geti SDK v2.5.0

What's Changed

• Introduce delete_dataset method by @igor-davidyuk in #489
• Update the download_all method for ImageClient to download specific dataset by @rajeshgangireddy in #484
• Update media info schema to include extension by @igor-davidyuk in #494
• Project access refactoring by @igor-davidyuk in #481
• Images upload from list optimization by @igor-davidyuk in #495
• Add readthedocs configuration file by @ljcornel in #497
• Fix bug with label name mapping for hierarchical classification models by @ljcornel in #498
• Introduce new Anomaly task by @igor-davidyuk in #487
• Implement get_training_dataset method in DatasetClient by @ljcornel in #492
• Fix anomaly task title by @igor-davidyuk in #499
• Anomaly tasks backwards compatibility by @igor-davidyuk in #502
• Fix anomaly demo test by @igor-davidyuk in #501
• Introduce project removal delay for short tests by @igor-davidyuk in #500
• Update datumaro requirement from ==1.8.* to ==1.9.* in /requirements by @dependabot in #496
• Update pytest requirement from ==8.2.* to ==8.3.* in /requirements by @dependabot in #482
• Fix organization id fetching by @ljcornel in #504
• Update ModelGroup data model by @ljcornel in #506
• Update Model data model by @ljcornel in #508
• Update pillow requirement from ==10.4.* to ==11.0.* in /requirements by @dependabot in #507
• Fix label assignment for deployed models by @ljcornel in #505
• Fix anomaly deployed models by @ljcornel in #510
• Keep label ID's intact on saving project deployment by @ljcornel in #511
• Fix anomaly wrapper configuration by @ljcornel in #513
• Fix empty label assignment for instance segmentation models by @ljcornel in #514
• Project deployment by name by @Daankrol in #516
• Bug Fix - Notebooks using project_name instead of project for a few methods by @rajeshgangireddy in #517
• Wait for job to appear in job queue before checking job status by @Daankrol in #518
• Fix 403 on delete project when running all cells in one go. by @Daankrol in #519
• Allow asynchronous annotation upload by @Daankrol in #509
• Fix errors in notebooks when running from E2E tests by @Daankrol in #522

New Contributors

• @rajeshgangireddy made their first contribution in #484
• @Daankrol made their first contribution in #516

Full Changelog: v2.3.0...v2.5.0

v2.3.0

What's Changed

• Add method to purge models by @igor-davidyuk in #468
• Fix visualization in 008 example notebook by @igor-davidyuk in #472
• Update the Model schema to include the total_disk_size field by @igor-davidyuk in #471
• Add support for Credit System API by @igor-davidyuk in #473
• Implement Platform version check by @igor-davidyuk in #474
• Allow Credit System-less servers to authenticate with a token by @igor-davidyuk in #476
• Empty label visualization fix by @igor-davidyuk in #480

Full Changelog: v2.2.0...v2.3.0

Geti SDK v2.2.0

What's Changed

• Add description attribute to the job class by @igor-davidyuk in #448
• Project/Dataset export import API alignment by @igor-davidyuk in #446
• Create geti_sdk.import_export docs page by @igor-davidyuk in #451
• Introduce data_model purge_info field by @igor-davidyuk in #453
• Disable certificate validation for mvtec dataset download by @ljcornel in #452
• Use proper project name in notebook 008 by @ljcornel in #454
• Update pillow requirement from ==10.3.* to ==10.4.* in /requirements by @dependabot in #449
• Create workflow to build wheels for different OS/CPU architecture combinations by @ljcornel in #456
• Fix headers in HttpRequest PostInferenceAction by @ljcornel in #457
• Correctly set headers in HttpRequest PostInferenceAction by @ljcornel in #458
• Expose Visualizer class by @igor-davidyuk in #426
• Job and Model schema update by @igor-davidyuk in #459
• Use json instead of data parameter in HttpRequestAction by @ljcornel in #460
• Remove workflows that require a self-hosted runner by @ljcornel in #461
• Fix 005_modify_image.ipynb image usage by @igor-davidyuk in #463
• Fix permissions for wheel build workflow and remove record_cassettes workflow by @ljcornel in #462
• Fix delete_project method by @igor-davidyuk in #466
• Add a Custom Dataset upload example by @igor-davidyuk in #464
• Add Job Cost info to job data scheme by @igor-davidyuk in #467

Full Changelog: releases/v2.1.0...1234

Geti SDK v2.1.0

What's Changed

• Saliency map visualization by @igor-davidyuk in #424
• Add a model management notebook by @igor-davidyuk in #419
• Add support for asynchronous inference in deployments by @ljcornel in #421
• Bump imageio-ffmpeg from 0.4.9 to 0.5.1 in /requirements by @dependabot in #430
• Saliency map fix by @igor-davidyuk in #427
• Fix rotated detectionsaliency map processing by @igor-davidyuk in #431
• Handle projects with duplicate names in the ProjectClient by @ljcornel in #425
• Update opencv-python requirement from ==4.9.* to ==4.10.* in /requirements by @dependabot in #428
• Update nightly test conda env directory by @ljcornel in #432
• Fix is_geti version check by @igor-davidyuk in #434
• Fix build and publish workflow for internal pypi by @ljcornel in #433
• Add check to make sure project is trained in notebook 012 by @ljcornel in #439
• Update flake8 requirement from ==7.0.* to ==7.1.* in /requirements by @dependabot in #440
• Fix removal of empty strings in json data for POSTing by @ljcornel in #442
• Include model_group_id in Model serialization and deserialization by @ljcornel in #443
• Include model_group_id upon model deployment in DeploymentClient by @ljcornel in #444
• Small fixes for testing client and test result by @ljcornel in #438
• Run pre-merge tests on PR to release branch by @ljcornel in #436
• Fix model_group_id assignment in DeployedModel deserialization by @ljcornel in #445
• Fix bug in AsyncVideoProcessor by @ljcornel in #441

Full Changelog: v2.0.1...v2.1.0

Geti SDK v2.0.2

What's Changed

• Fix is_geti version check (#434) by @igor-davidyuk in #435

Full Changelog: v2.0.1...v2.0.2

Geti SDK v2.0.1

What's Changed

• Add retry mechanism to better handle ConnectionError by @ljcornel in #420
• Undo changes to video tempfile handling, register atexit handler for tempfile deletion by @ljcornel in #423

Full Changelog: v2.0.0...v2.0.1

Geti SDK v2.0.0

New features

This release introduces a new feature related to model deployment: post-inference hooks! A post-inference hook can be added to any Deployment, and will be executed after every inference request (i.e. every call to deployment.infer()). The hooks allow you to define specific actions to take under certain conditions. For example, a hook could implement the following behaviour:
If the confidence level of one of the predictions for the image is less than 20%, then upload the image to the Intel® Geti™ project in which the model was trained.

This could be useful for improving your model with a next training round, because including such 'low confidence images' in the training dataset might help to improve model accuracy.
Additional examples of post-inference hooks, and instructions for configuring them, can be found in the newly added notebook 012 in this repository.

Breaking changes

This major release of the Intel® Geti™ SDK breaks backwards compatibility with Intel® Geti™ servers of version v1.14 and below. Please make sure that your Intel® Geti™ server is updated to the latest version of the Intel® Geti™ platform, to prevent compatibility issues.

What's Changed

• Update Video data model with annotation statistics by @ljcornel in #391
• Record Legacy and Develop cassette in separate steps instead of pipelines by @igor-davidyuk in #387
• Remove dependency on OTX by @igor-davidyuk in #393
• Add model_storage_id to models when fetching model group by @ljcornel in #399
• Add Python 3.11 support; Deprecate Python 3.8 by @igor-davidyuk in #398
• Add last_annotator_id field to media data model by @igor-davidyuk in #403
• Documentation Update 2.0 by @igor-davidyuk in #402
• End support for Platforms versions lower than 1.15 by @igor-davidyuk in #397
• Bump imageio-ffmpeg from 0.4.8 to 0.4.9 in /requirements by @dependabot in #404
• Update python version in github workflows by @ljcornel in #405
• Update requirement for tqdm to >=4.66.3 by @ljcornel in #407
• Add PostInferenceHook feature initial implementation and notebook by @ljcornel in #355
• [StepSecurity] Apply security best practices by @step-security-bot in #408
• Define permissions on job level for cassette record workflow by @ljcornel in #409
• Update pytest-recording requirement from ==0.12.* to ==0.13.* in /requirements by @dependabot in #406
• Update pytest requirement from ==7.4.* to ==8.2.* in /requirements by @dependabot in #401
• Update cryptography requirement to >=42.0.7 by @ljcornel in #410
• Update requirement for joblib to >=1.2.0 by @ljcornel in #411
• Update certifi requirement to >=2024.2.2 by @ljcornel in #412
• Update ipython requirement from ==8.12.* to ==8.18.* in /requirements by @dependabot in #400
• Update vcrpy requirement from ==4.4.* to ==6.0.* in /requirements by @dependabot in #318
• Remove dead code in training client by @ljcornel in #413
• Update requests requirement from ==2.31.* to ==2.32.* in /requirements by @dependabot in #415
• Update pytest-html requirement from ==3.2.* to ==4.1.* in /requirements by @dependabot in #418
• Update pytest-cov requirement from ==4.1.* to ==5.0.* in /requirements by @dependabot in #416
• Add initial support for Tiling inference [synchronous mode only] by @ljcornel in #414

New Contributors

• @step-security-bot made their first contribution in #408

Full Changelog: v1.16.1...v2.0.0

Geti SDK v1.16.1

What's Changed

• Add default_workspace to possible default workspace names by @igor-davidyuk in #394

Full Changelog: v1.16.0...v1.16.1

Geti SDK v1.16.0

What's Changed

• Decouple visualizer for OTX by @igor-davidyuk in #356
• Update data models to account for REST API changes in Geti v1.16 by @ljcornel in #357
• Fix organization retrieval URL with PAT by @ljcornel in #361
• TEST: add VCR recording workflow by @igor-davidyuk in #360
• Fix job id retrieval in optimize_model method by @ljcornel in #364
• Update job datamodel by @ljcornel in #362

Full Changelog: v1.15.0...v1.16.0

Geti SDK v1.15.0

This release makes the SDK compatible for Intel® Geti™ v1.15. The majority of the changes is focused on that, but below is a list of other changes that are worth mentioning.

Release Highlights

• Add compatibility for Intel® Geti™ v1.15. Note that we maintain backwards compatibility for models created in Intel® Geti™ v1.8: This means that the SDK can run inference on deployments from both the latest Intel® Geti™ on-prem and Intel® Geti™ cloud-based release.
• Add benchmarking functionality for deployments through the Benchmarker class, as well as a new notebook to demonstrate this feature.
• Improve the job monitoring feature, job progression is now displayed via progress bars.
• Image and annotation upload and download now uses multithreading, greatly speeding up the process.
• A new notebook demonstrating an end-to-end workflow, from model creation to deployment, was added to notebooks/use_cases.
• Multiple dependency updates, minor bug fixes and documentation improvements.

What's Changed

• Add organization ID to session when appropriate by @ljcornel in #294
• Change job monitor functions to use progress bars by @ljcornel in #295
• Decrease verbosity for annotation download by @ljcornel in #296
• Improve model selection mechanism by @ljcornel in #297
• Add Benchmarker class and demo notebook for inference throughput benchmarking of Geti models by @ljcornel in #300
• Supported algorithms workaround by @ljcornel in #303
• Workaround for getting supported algorithms by @ljcornel in #305
• Update README.md by @operepel in #307
• Fix organization ID when uploading annotations by @ljcornel in #306
• Mention git lfs in test docs by @igor-davidyuk in #308
• Fix inference endpoints and job data model by @ljcornel in #309
• Improve handling of supported algorithms by @ljcornel in #310
• Fix TaskAnnotationState and Optimization JobType by @ljcornel in #312
• Test benchmarker by @igor-davidyuk in #311
• Pin otx to v1.4.3 by @ljcornel in #313
• Remove snyk scan from security workflow by @ljcornel in #314
• Minor bug fixes related to deployment by @ljcornel in #315
• Inference results visual comparison by @igor-davidyuk in #316
• Add model_activated field to ModelMetadata by @ljcornel in #320
• Update Pillow dependency to Pillow==10.2.* by @ljcornel in #322
• Add shields.io badges to readme by @ljcornel in #321
• Relax time requirement for demo projects nightly tests by @ljcornel in #325
• Multithread image upload and download by @igor-davidyuk in #323
• Handle the new authorization mechanism and organization IDs by @ljcornel in #326
• Improve robustness of job monitoring by @ljcornel in #327
• Introduce method to switch active model by @igor-davidyuk in #328
• Fix issue with undefined label color for prediction by @ljcornel in #330
• Enable pre-merge tests on Windows and MacOS by @igor-davidyuk in #331
• Update compatibility table in README by @ljcornel in #332
• Add compatibility for anomaly model deployment for Geti v1.13 and up by @ljcornel in #333
• Correctly handle errors in new dex authentication mechanism by @ljcornel in #336
• Handle the new personal access tokens for Geti v1.15 and up by @ljcornel in #335
• Fix proxies issue in GetiSession by @igor-davidyuk in #334
• Modified 'Smart Cities' notebook by @ljcornel in #304
• Fix annotation kind comparison in plot_helpers by @ljcornel in #337
• Update flake8 requirement from ==6.0.* to ==7.0.* in /requirements by @dependabot in #302
• Handle org ID with personal access tokens for Intel Geti cloud instance by @ljcornel in #338
• Bump orjson from 3.9.2 to 3.9.15 in /requirements by @dependabot in #344
• Update pytest-env requirement from ==1.0.* to ==1.1.* in /requirements by @dependabot in #341
• Update datumaro requirement from ==1.4.* to ==1.5.* in /requirements by @dependabot in #347
• geti_session - remove unused proxy kwargs by @igor-davidyuk in #346
• CVS-117852 Fix bandit issues by @ljcornel in #348
• Make datasets optional in Project data model by @ljcornel in #339
• Fix maskrcnn postprocessing for geti v1.15 by @ljcornel in #350
• Restrict default top-level permissions for all workflows by @ljcornel in #349
• Add openssf scorecard code scanning by @ljcornel in #351
• OTX decoupling part 1: inference postprocessing by @ljcornel in #352
• OTX decoupling: deployment by @igor-davidyuk in #345

New Contributors

• @operepel made their first contribution in #307
• @igor-davidyuk made their first contribution in #308

Full Changelog: v.1.8.2...v1.15.0