GitLab

* [proto] Added few transforms tests, part 1 (#6262)

* Added supported/unsupported data checks in the tests for cutmix/mixup

* Added RandomRotation, RandomAffine transforms tests

* Added tests for RandomZoomOut, Pad

* Update test_prototype_transforms.py

* Added RandomCrop transform and tests (#6271)

* [proto] Added GaussianBlur transform and tests (#6273)

* Added GaussianBlur transform and tests

* Fixing code format

* Copied correctness test

* [proto] Added random color transforms and tests (#6275)

* Added random color transforms and tests

* Disable smoke test for RandomSolarize, RandomAdjustSharpness

* Added RandomPerspective and tests (#6284)

- replaced real image creation by mocks for other tests

* Added more functional tests (#6285)

* [proto] Added elastic transform and tests (#6295)

* WIP [proto] Added functional elastic transform with tests

* Added more functional tests

* WIP on elastic op

* Added elastic transform and tests

* Added tests

* Added tests for ElasticTransform

* Try to format code as in https://github.com/pytorch/vision/pull/5106



* Fixed bug in affine get_params test

* Implemented RandomErase on PIL input as fallback to tensors (#6309)

Added tests

* Added image_size computation for BoundingBox.rotate if expand (#6319)

* Added image_size computation for BoundingBox.rotate if expand

* Added tests

* Added erase_image_pil and eager/jit erase_image_tensor test (#6320)

* Updates according to the review
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Refactor tests

* Remove tol, fix comments

* Add tolerance only where necessary

* Add tolerance only where necessary

* Add tolerance only where necessary

* Refactor to adapt suggestions

* Refactor and add nits

* Refactor box area

* Refactor to one file

* Adapt almost all except area

* final update

* Tighten for jit

* Refactor slightly

* Fix tests
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* [proto] Added missing decorator for center_crop_segmentation_mask tests

* Added center_crop_image_tensor

* upgrade usort to

* Also update black

* Actually use 1.0.2

* Apply pre-commit
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Another attempt to add transforms

* Fixed padding type hint

* Fixed fill arg for pad and rotate, affine

* code formatting and type hints for affine transformation

* Fixed flake8

* Updated tests to save and load transforms

* Fixed code formatting issue

* Fixed jit loading issue

* Restored fill default value to None
Updated code according to the review

* Added tests for rotation, affine and zoom transforms

* Put back commented code

* Random erase bypass boxes and masks
Go back with if-return/elif-return/else-return

* Fixed acceptable and non-acceptable types for Cutmix/Mixup

* Updated conditions for _BaseMixupCutmix

* Updated tests to save and load transforms

* Fixed code formatting issue

* Added mid-level ops and feature-based ops

* Fixing deadlock in dataloader with circular imports

* Added non-scalar fill support workaround for pad

* Removed comments

* int/float support for fill in pad op

* Updated type hints and removed bypass option from mid-level methods

* Minor nit fixes

* Adding MViT v2 architecture (#6105)

* Adding mvitv2 architecture

* Fixing memory issues on tests and minor refactorings.

* Adding input validation

* Adding docs and minor refactoring

* Add `min_temporal_size` in the supported meta-data.

* Switch Tuple[int, int, int] with List[int] to support easier the 2D case

* Adding more docs and references

* Change naming conventions of classes to follow the same pattern as MobileNetV3

* Fix test breakage.

* Update todos

* Performance optimizations.

* Add support to MViT v1 (#6179)

* Switch implementation to v1 variant.

* Fix docs

* Adding back a v2 pseudovariant

* Changing the way the network are configured.

* Temporarily removing v2

* Adding weights.

* Expand _squeeze/_unsqueeze to support arbitrary dims.

* Update references script.

* Fix tests.

* Fixing frames and preprocessing.

* Fix std/mean values in transforms.

* Add permanent Dropout and update the weights.

* Update accuracies.

* Fix documentation

* Remove unnecessary expected file.

* Skip big model test

* Rewrite the configuration logic to reduce LOC.

* Fix mypy

* Added base tests for rotate_image_tensor

* Updated resize_image_tensor API and tests and fixed a bug with max_size

* Refactored and modified private api for resize functional op

* Fixed failures

* More updates

* Updated proto functional op: resize_image_*

* Added max_size arg to resize_bounding_box and updated basic tests

* Update functional.py

* Reverted fill/center order for rotate
Other nits

* Refactored and modified private api for resize functional op

* Fixed failures

* More updates

* Fixed flake8

* Add rough raft-stereo implementation on prototype/models

* Add standard raft_stereo builder, and modify context_encoder to be more similar with original implementation

* Follow original implementation on pre-convolve context

* Fix to make sure we can load original implementation weight and got same output

* reusing component from raft

* Make the raft_stereo_fast able to load original weight implementation

* Format with ufmt and update some comment

* Use raft FlowHead

* clean up comments

* Remove unnecessary import and use ufmt format

* Add __all__ and more docs for RaftStereo class

* Only accept param and not module for raft stereo builder

* Cleanup comment

* Adding typing to raft_stereo

* Update some of raft code and reuse on raft stereo

* Use bool instead of int

* Make standard raft_stereo model jit scriptable

* Make the function _make_out_layer using boolean with_block and init the block_layer with identity

* Separate corr_block into two modules for pyramid and building corr features

* Use tuple if input is not variable size, also remove default value if using List

* Format using ufmt and update ConvGRU to not inherit from raft in order to satisfy both jit script and mypy

* Change RaftStereo docs input type

* Ufmt format raft

* revert back convgru to see mypy errors, add test for jit and fx, make the model fx compatible

* ufmt format

* Specify device for new tensor, dont init module then overwrite and put if-else instead

* Ignore mypy problem on override, put back num_iters on forward

* Revert some effort to make it fx compatible but unnecessary now

* refactor code and remove num_iters from RaftStereo constructor

* Change to raft_stereo_realtime, and specify device directly for tensor creation

* Add description for raft_stereo_realtime

* Update the test for raft_stereo

* Fix raft stereo prototype test to properly test jit script

* Ufmt format

* Test against expected file, change name from raft_stereo to raft_stereo_builder to prevent import error

* Revert __init__.py changes

* Add default value for non-list param on model builder

* Add checking on out_with_block length, add more docs on the encoder

* Use base instead of basic since it is more commonly used

* rename expect file to base as well

* rename on test

* Revert the revert of __init__.py, also revert the adding default value to _raft_stereo to follow the standard pattern

* ufmt format __init__.py

* Added elastic augment

* ufmt formatting

* updated comments

* fixed circular dependency issue and bare except error

* Fixed three type checking errors in functional_tensor.py

* ufmt formatted

* changed elastic_deformation to a more common implementation

Implementation uses alpha and sigma to control strength and smoothness of the displacement vectors in elastic_deformation instead of control_point_spacings and sigma.

* ufmt formatting

* Some performance updates

Put random offset vectors to device before gaussian_blur is applied speeds it up 3-fold.

* fixed type error

* fixed again a type error

* Update torchvision/transforms/functional_tensor.py
Co-authored-by: vfdev <vfdev.5@gmail.com>

* Added some requested changes

- pil image support similar to GaussianBlur
- changed interpolation arg to InterpolationMode
- added a wrapper in torchvision.transforms.functional.py that gets called by the class in transforms.py
-renamed it to ElasticTransform
- handled sigma = 0 case

* added img docstring

* added some tests

* Updated tests and the code

* Added the requested changes to the arguments of F.elastic_transform

Added random_state and displacement as arguments to F.elastic_transform

* fixed the type error

* Fixed tests and docs

* implemented requested changes

Changes:
1) alpha AND sigma OR displacement must be given as arguments to transforms.functional_tensor.elastic_transform instead of alpha AND sigma AND displacement
2) displacements are accepted in transforms.functional.elastic_transform as np.array and torch.Tensor instead of only accepting torch.Tensor

* ufmt formatting

* trochscript error resolved

replaced torch.from_numpy() to torch.Tensor() to make it compatible to torchscript

* revert to torch.from_numpy()

* updated argument checks and errors

- In F.elastic_transform added check to see if both user inputs img and displacement are either of type PIL Image and ndarray or both of type tensor.
- In F_t.elastic_transform added check if alpha and sigma are None if displacement is given or vice versa.

* fixed seed error

changed torch.seed to torch.manual_seed in F_t.elastic_transform

* Reverted displacement type and other cosmetics

* Other minor improvements

* changed gaussian_blur filter size

changed gaussian_blur filter size
from
4 * int(sigma) + 1
to
int(8 * sigma + 1)
to make it consistent with ernestums implementation

* resolved merge error

* Revert "resolved merge error"

This reverts commit 6a4a4e74ff4d078e2c2753d359185f9a81c415d0.

* resolve merge error

* ufmt formatted

* ufmt formated once again..

* fixed unsupported operand error

* Update API and removed random_state from functional part

* Added default values

* Added ElasticTransform to gallery and updated the docstring

* Updated gallery and added _log_api_usage_once
BTW, matplotlib.pylab is deprecated

* Updated gallery transforms code

* Updates according to review
Co-authored-by: vfdev <vfdev.5@gmail.com>

* Update _pil_constants.py

* Update _pil_constants.py

* Fix flake8

* Fixed two related warnings in tests

* switch dir with hasattr
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Fix the issue :)

* Intellij vs ufmt battle

* remove .item()

* Added tests for other padding modes

* Fixed expected mask dtype

* Applied comments from review

* add tests for F.pad_bounding_box

* Added correctness tests for pad and reimplemented bbox op to keep dtype

* Update _geometry.py
Co-authored-by: vfdev <vfdev.5@gmail.com>

Add .float() before .mean() on test_backbone_utils.py because .mean() dont accept integer dtype (#6090)

* Validate against expected files on videos

* Plus tests for autocast

* feat: add functional center crop on mask

* test: add correctness center crop with random segmentation mask

* test: improvements

* test: improvements

* Apply suggestions from code review
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Federico Pozzi <federico.pozzi@argo.vision>
Co-authored-by: vfdev <vfdev.5@gmail.com>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Remove `(N, T, H, W, C) => (N, T, C, H, W)` conversion on presets

* Update docs.

* Fix the tests

* Use `output_format` for `read_video()`

* Use `output_format` for `Kinetics()`

* Adding input descriptions on presets

Throw warning for empty masks or box tensors on draw_segmentation_masks and draw_bounding_boxes (#5857)

* Fixing the IndexError in draw_segmentation_masks

* fixing the bug on draw_bounding_boxes

* Changing fstring to normal string

* Removing unecessary conversion

* Adding test for the change

* Adding a test for draw seqmentation mask

* Fixing small mistake

* Fixing an error in the tests

* removing useless imports

* ufmt
Co-authored-by: LEGRAND Matthieu <legrand.ma@chu-toulouse.fr>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Add swin on hubconfig.

* Add swin b/s in the `slow_models` list.

* add swin_s and swin_b variants

* fix swin_b params

* fix n parameters and acc numbers

* adding missing acc numbers

* apply ufmt

* Updating `_docs` to reflect training recipe

* Fix exted for swin_b
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Classif models

* Detection

* Segmentation

* quantization

* Video

* optical flow

* tests

* Fix docs

* Fix Video dataset

* Consistency for RAFT dataset names

* use ImageNet-1K

* Use COCO-val2017-VOC-labels for segmentation

* formatting

* Adding docs for quantized models.

* Adding docs for video models.

* Adding docs for segmentation models.

* Adding docs for optical flow models.

* Adding docs for detection models.

* Fix typo.

* Make changes from code-review.

* Improving the auto-gen doc.

* Adding details for AlexNet, ConvNext, DenseNet, EfficientNets, GoogLeNet and InceptionV3.

* Fixing location of `_docs`

* Adding docs in the remaining classification models.

* Fix linter

* simplify OnlineResource.load

* [PoC] merge mock data preparation and loading

* Revert "cache mock data based on config"

This reverts commit 5ed6eedef74865e0baa746a375d5ec1f0ab1bde7.

* Revert "[PoC] merge mock data preparation and loading"

This reverts commit d62747962f9ed6a7b0b80849e7c971efabb5d3da.

* remove preprocess returning a new path in favor of querying twice

* address test comments

* clarify comment

* mypy

* use builtin decompress utility

* merge mock data preparation and loading

* address comments

* fix extra file creation

* remove tmp folder

* inline images meta creation in coco mock data

* Requested here https://github.com/pytorch/vision/pull/5898#discussion_r864765799

.

* Fix tests

* ufmt, not black
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* rely on patched datasets home rather than passing it around

* add comment

* POC

* Update torchvision/models/resnet.py
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Apply suggestions from code review
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Fix tests

* ufmt

* Remove useless docstring
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* [FEAT] Add distance IoU and distance IoU loss + some tests (WIP for tests).

* [FIX] Remove URL from docstring + remove assert since it causes a big performance drop.

* [FIX] eps isn't None.

* [TEST] Update existing box dIoU test + add dIoU loss tests (inspired from cIoU ones).

* [ENH] Some pre-commit fixes + remove print + mypy.

* [ENH] Pass the device in the assertion for the dIoU loss test.

* [FIX] Remove type hints from the dIoU box test.

* [ENH] Refactor box and loss for dIoU functions + fix half tests.

* [FIX] Precommits fix.

* [ENH] Some improvement for the distance IoU tests thanks to code review.

* [ENH] Upcast in distance boxes computation to avoid overflow.

* [ENH] Revert the refactor of distance IoU loss back since it introduced a bug and can be slow.

* Precommit fix.

* [FIX] Few changes introduced by merge conflict.

* Add code reference

* Fix test
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* [proto] Added `center_crop_bounding_box` functional op

* Fixed mypy issue

* Added one more test case

* More test cases

* Added functional `perspective_bounding_box`/`perspective_segmentation_mask` ops

* Added more comments and added a code to assert denom != 0

* Put larger r/a tolerence when matching bboxes

* Update transforms for PIL deprecation

* Changes agreed at pytorch/vision#5898

* black, sort constants, version check

* Format tests

* Square brackets

* Update torchvision/transforms/_pil_constants.py
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Add resnext101_64x4d model definition

* Add test for resnext101_64x4d

* Add resnext101_64x4d weight

* Update checkpoint to use EMA weigth

* Add quantization model signature for resnext101_64x4d

* Fix class name and update accuracy using 1 gpu and batch_size=1

* Apply ufmt

* Update the quantized weight and accuracy that we still keep the training log

* Add quantized expect file

* Update docs and fix acc1

* Add recipe for quantized to PR

* Update models.rst