Release v1.5.0

New features

• Enable configurable confidence threshold for otx eval and export (#2388)
• Add YOLOX variants as new object detector models (#2402)
• Enable FeatureVectorHook to support action tasks (#2408)
• Add ONNX metadata to detection, instance segmantation, and segmentation models (#2418)
• Add a new feature to configure input size (#2420)
• Introduce the OTXSampler and AdaptiveRepeatDataHook to achieve faster training at the small data regime (#2428)
• Add a new object detector Lite-DINO (#2457)
• Add Semi-SL Mean Teacher algorithm for Instance Segmentation task (#2444)
• Official supports for YOLOX-X, YOLOX-L, YOLOX-S, ResNeXt101-ATSS (#2485)
• Add new argument to track resource usage in train command (#2500)
• Add Self-SL for semantic segmentation of SegNext families (#2215)
• Adapt input size automatically based on dataset statistics (#2499)

Enhancements

• Refine input data in-memory caching (#2416)
• Adapt timeout value of initialization for distributed training (#2422)
• Optimize data loading by merging load & resize operations w/ caching support for cls/det/iseg/sseg (#2438, #2453, #2460)
• Support torch==2.0.1 (#2465)
• Set "Auto" as default input size mode (#2515)

Bug fixes

• Fix F1 auto-threshold to choose best largest confidence (#2371)
• Fix a performance drop while training EfficientNetV2 with multi-GPU (#2398)
• Fix bug that auto adaptive batch size raises an error if CUDA isn't available (#2410)
• Fix sampler degradation issue (#2482)
• Fix a bug that fp16 key error is raised when training without GPU (#2501)
• Fix bug that auto batch size doesn't consider distributed training (#2533)
• Fix auto input size mismatch in eval & export (#2530)
• Fix the CustomNonLinearClsHead when the batch_size is set to 1 (#2571)
• Fix IBLoss enablement with DeiT-Tiny when class incremental training (#2594)
• Fix XAI algorithm for Detection (#2617)

Full Changelog: 1.4.4...1.5.0

Release v1.4.4

What's Changed

• Update ModelAPI configuration by @sovrasov in #2564
• Add Anomaly modelAPI changes to releases/1.4.0 by @ashwinvaidya17 in #2563
• Update for 1.4.4rc1 by @yunchu in #2572
• Bump datumaro version to 1.5.1rc3 by @yunchu in #2587
• Add label ids to anomaly OpenVINO model xml by @ashwinvaidya17 in #2590
• Fix DeiT-Tiny regression test for release/1.4.0 by @wonjuleee in #2595
• Fix mmcls bug not wrapping model in DataParallel on CPUs by @goodsong81 in #2601
• Fix h-label loss normalization issue w/ exclusive label group of singe label by @goodsong81 in #2604
• Boost up Image numpy accessing speed through PIL by @wonjuleee in #2586
• Fix division by zero in class incremental learning for classification by @goodsong81 in #2606
• Fix XAI algorithm for Detection by @GalyaZalesskaya in #2609
• Tighten dependency constraint only adapting latest patches by @yunchu in #2607
• Add metadata to optimized model by @ashwinvaidya17 in #2618
• Update dependency constraint by @yunchu in #2622
• Fix h-label bug of missing parent labels in output by @goodsong81 in #2626
• Update publish workflow by @yunchu in #2625
• Update publish workflow for tag checking by @yunchu in #2632
• Update release note and readme by @yunchu in #2637
• Make max_num_detections configurable by @goodsong81 in #2647
• Fix CPU training issue on non-CUDA system by @goodsong81 in #2655
• Update MAPI version by @sovrasov in #2730
• Update dependency for exportable code by @yunchu in #2732

Full Changelog: 1.4.3...1.4.4

Release v1.4.4rc1

What's Changed

• Update ModelAPI configuration by @sovrasov in #2564
• Add Anomaly modelAPI changes to releases/1.4.0 by @ashwinvaidya17 in #2563
• Update for 1.4.4rc1 by @yunchu in #2572
• Bump datumaro version to 1.5.1rc3 by @yunchu in #2587

Full Changelog: 1.4.3...1.4.4rc1

Release v1.5.0rc1

What's Changed

• POT->PTQ renaming in tests by @sovrasov in #2319
• Update the version of the openvino badge by @sungmanc in #2329
• Tiling doc for develop by @eugene123tw in #2334
• Relax reference saliency values check by @negvet in #2338
• Call otx.cli before the TaskType is imported by @harimkang in #2342
• added fix-cpu numbers option by @yunchu in #2340
• Fix tiling unittest and random max select logic by @eugene123tw in #2336
• Update incorrect license header by @yunchu in #2352
• Fixing XAI tests by @negvet in #2357
• Change IR scale factor to 1.0 if tile size is too big by @eugene123tw in #2337
• Add model category attributes to model template by @goodsong81 in #2362
• Fix F1 auto-threshold to choose best largest confidence by @goodsong81 in #2371
• Add model type to deit ptq config by @sovrasov in #2379
• Add description & remove unused configurable parameters by @goodsong81 in #2382
• Add onnx metadata to classification models by @sovrasov in #2378
• Update exportable code README (to develop) by @GalyaZalesskaya in #2389
• Rename SAMImageClassifier -> CustomImageClassifier by @goodsong81 in #2384
• Merge back 1.4.0 to develop by @yunchu in #2387
• Update version string of the OTX by @yunchu in #2393
• [develop] XAI: Return saliency maps for Mask RCNN IR async infer by @GalyaZalesskaya in #2391
• Enable deterministic for anomalib and visual prompting by @sungchul2 in #2361
• Fix a performance drop while training EfficientNetV2 with multi-GPU by @eunwoosh in #2398
• Change POT to PTQ in test code by @eunwoosh in #2401
• Enable configurable confidence threshold for ov export and inference by @jaegukhyun in #2388
• Fix e2e by @jaegukhyun in #2404
• Add YOLOX variants to OTX by @jaegukhyun in #2402
• Enable FeatureVectorHook for action tasks by @jaegukhyun in #2408
• Update Anomaly OpenVINO export for ModelAPI by @ashwinvaidya17 in #2400
• Fix bug that auto adaptive batch size raises an error if CUDA isn't available by @eunwoosh in #2410
• Check sal maps for async infer for M-RCNN by @negvet in #2396
• Remove data cfg in detection training recipe by @jaegukhyun in #2409
• Switch to MAPI in classification and detection inference tasks by @sovrasov in #2390
• Mergeback 1.4.1 to develop by @yunchu in #2413
• Enable ViTReciproCAM for VisionTransformer backbone (cls task) by @negvet in #2403
• Add onnx meta for sseg, iseg, and detection models by @sovrasov in #2418
• Reduce amount of SSeg PTQ e2e tests by @sovrasov in #2414
• Refactor instance segmentation and tiling inference by @sovrasov in #2415
• Refine input data in-memory caching by @goodsong81 in #2416
• Configurable input size by @eunwoosh in #2420
• Adpat timeout value of torch.dist.init_process_group depending on elapsed time by @eunwoosh in #2422
• Refactor semantic segmentation inference to use MAPI by @sovrasov in #2421
• Update FQ refs for iseg models by @sovrasov in #2426
• Limit range of PTQ parameters by @sovrasov in #2425
• Refactor configurer - Phase1 by @jaegukhyun in #2424
• Upgrade ModelAPI by @sovrasov in #2429
• Move SegNext models from experimental to custom models by @kprokofi in #2312
• Refactor configurer - Phase 2 by @jaegukhyun in #2430
• Introduce AdaptiveRepeatDataHook for the classification by @sungmanc in #2428
• Write a docs and cover additional case about configurable input size by @eunwoosh in #2433
• Fix E2E, Regression test failures and edit the inproper log message by @sungmanc in #2436
• Enable multi-gpus for weekly by @yunchu in #2364
• Add MaskRCNN SwinTransformer-T by @eugene123tw in #2423
• Fix i-seg e2e test code by @eunwoosh in #2446
• Stabilize multi GPU training by @eunwoosh in #2443
• Merge load / resize / cache to optimize data loading efficiency for classification by @goodsong81 in #2438
• Refactor configurer - Phase 3 by @jaegukhyun in #2442
• Merge load / resize / cache to optimize data loading efficiency for detection & instance segmentation by @goodsong81 in #2453
• Align resume batch size with train batch size in i-seg e2e test by @eunwoosh in #2458
• Merge load / resize / cache to optimize data loading efficiency for semantic segmentation by @goodsong81 in #2460
• Apply load+resize cache to validation pipelines by @goodsong81 in #2463
• Add a new object detector, Lite-DINO by @jaegukhyun in #2457
• Bump datumaro version by @yunchu in #2466
• Set tox version constraint by @yunchu in #2471
• Add utilization of the unlabeled subset in PTQ optimization by @kprokofi in #2468
• Remove unnecessary workflows by @yunchu in #2474
• Add unit tests for DINO and Lite-DINO by @jaegukhyun in #2475
• Mergeback 1.4.2rc2 by @yunchu in #2464
• Added quantization configs for SegNext T S B by @nikita-savelyevv in #2477
• Update regression criteria for anomaly task by @yunchu in #2476
• Fix some issues on cls regression tests by @yunchu in #2478
• Adjust PTQ fq target by @chuneuny-emily in #2480
• Add Semi-SL Instance Segmentation by @kprokofi in #2444
• Fix some bugs for reg tests by @yunchu in #2481
• Fix sampler degradation issue by @sungmanc in #2482
• Make MaskRCNN-SwinT from experimental to production by @eugene123tw in #2454
• Remove mpa legacy by @jaegukhyun in #2486
• Fix some regression test issues for segmentation by @yunchu in #2488
• Re-organization of configurer by @jaegukhyun in #2484
• Model revamp for Get...

Release v1.4.3

Enhancements

• Re introduce adaptive scheduling for training (#2541)

Full Changelog: 1.4.2...1.4.3

Release v1.4.3rc1

What's Changed

• Re introduce adaptive scheduling for training by @jaegukhyun in #2541
• Update for release 1.4.3rc1 by @yunchu in #2542

Full Changelog: 1.4.2...1.4.3rc1

Release v1.4.2

Enhancements

• Add workaround for the incorrect meta info M-RCNN (used for XAI) (#2437)
• Add model category attributes to model template (for releases/1.4.0) (#2439)
• Update PTQ ignored scope for hrnet 18 mod2 (#2449)
• Modified fq numbers for lite HRNET (#2445)
• Prevent zero-sized saliency map in tiling if tile size is too big (#2452)
• Bump datumaro version (#2502)
• Add a dummy hierarchical config required by MAPI (#2483)
• Set tox version constraint (#2472)
• Update pot fq reference number (#2456)
• Upgrade nncf version to 2.6.0 (#2459)

Bug fixes

• Fix label list order for h-label classification (#2440)
• Fix OpenVINO inference for legacy models (#2450)
• Bug fix for albumentations (#2467)

Full Changelog: 1.4.1...1.4.2

Release v1.4.0rc5

What's Changed

• Bump datumaro version by @yunchu in #2502
• Upgrade nncf version for 1.4 release by @sovrasov in #2459
• Update version for release 1.4.2rc5 by @yunchu in #2507

Full Changelog: 1.4.2rc4...1.4.2rc5

Release v1.4.2rc4

What's Changed

• Bump datumaro version to 1.5.0rc0 by @yunchu in #2470
• Set tox version constraint by @yunchu in #2472
• Bug fix for albumentations by @ashwinvaidya17 in #2467
• Add a dummy hierarchical config required by MAPI by @sovrasov in #2483

Full Changelog: 1.4.2rc2...1.4.2rc4

Release v1.4.2rc2

What's Changed

• Update PTQ ignored scope for hrnet 18 mod2 by @sovrasov in #2449
• Fix OpenVINO inference for legacy models by @ashwinvaidya17 in #2450
• Update for 1.4.2rc2 by @yunchu in #2455
• Prevent zero-sized saliency map in tiling if tile size is too big by @sovrasov in #2452
• Update pot fq reference number by @yunchu in #2456

Full Changelog: 1.4.2rc1...1.4.2rc2