506 Commits

Author SHA1 Message Date
torzdf ac8206ed6e bugfix: Extract - rotation in detection 2023-06-15 18:28:37 +01:00
torzdf 02239d4183 locales: Train settings 2023-06-11 23:30:05 +01:00
torzdf a66214c58e locales: Add global extract settings 2023-06-08 16:45:18 +01:00
torzdf 0c1d1e3253 Bugfix: Color order on preview with learn_mask 2023-02-14 13:09:46 +00:00
torzdf 2028579756 Preview - Allow mask color/opacity configuration
- typing: lib.config
  - docs: lib.config
  - bugfix: popup_configure - keep assets on top
2023-02-13 15:03:34 +00:00
torzdf 8a5ee6d9bf Bugfix: Patch extract memory leak in batch mode 2023-02-10 15:52:34 +00:00
torzdf a1ef5edd39 tests:
- unit test: tools.alignments.media
  - Add mypy test
  - Typing fixes
2023-01-21 00:41:18 +00:00
torzdf 80f63280ca Bugfixes
- plugin.train.unbalanced - decoder b
  - gui.stats - crash when log folder deleted
  - setup.py - install gcc=12.1.0 on linux (scipy dep)
  - setup.py - don't test ROCm unless valid
  - typofix - cli_args extract singleprocess help
typing
  - lib.gui.analysis.event_reader
  - lib.gpu_stats.directml - make platform specific
unit tests
  - lib.gui.analysis.event_reader
2023-01-13 23:29:51 +00:00
torzdf 28cb2fcffb Unhide options for ROCm 2023-01-11 14:34:30 +00:00
torzdf 2f16f8aa3b Bugfix - Phaze-A, monkeypatch EfficientNet for MixedPrecision 2023-01-07 19:19:07 +00:00
torzdf dc94ed9cad DirectML Support (#1291)
* DML Support
  - Add requirements. Update setup.py. Update Windows Installer
  - Add 'directml' backend. Update backend supporting logic
  - GPUStats - Add GPU querying from Win SDK for DML backend
  - Windows PlaidML - Deprecation Warning
  - Remove deprecated '-d' training flag
  - Logging in DML functions
- Unittest
  - Windows Unit Test
  - lib.utils
2023-01-06 14:45:57 +00:00
torzdf d750af1e92 bump TF to version 2.10 2022-12-20 15:30:26 +00:00
torzdf 84ae6006bb Bugfix - TF Version check 2022-12-19 11:26:32 +00:00
torzdf 2faef58c5f Bugfix - Extract - re-align with identity filtering enabled 2022-11-22 17:34:44 +00:00
torzdf 9e2026f6fe Extract: Implement re-align/2nd pass
- implement configurable re-align function in extract
  - update locales + documentation
  - re-factor align._base and split to separate modules
  - move normalization method to plugin parent
  - bugfix: FAN use zeros for pre-processing crop
  - lint AlignedFilter
2022-11-22 00:37:29 +00:00
torzdf 113b7d7db4 Aligner filters bugfixes
- Skip roll check if disabled
  - Correct subfolder name for features check
2022-11-09 12:56:23 +00:00
torzdf c698f45a35 Aligner - Add feature position filter 2022-11-08 17:09:43 +00:00
torzdf cb8ec69789 Aligner updates
- Add filter re-feeds option
  - bugfix roll calculation
2022-11-08 12:10:56 +00:00
torzdf 0f5d2e887c Face-filter updates:
- Allow selecting folder as well as multiple images
  - Lower default threshold and update helptext
  - bugfix: detector error when using all aligned faces
  - Standardize output folder name
2022-11-04 18:43:55 +00:00
torzdf 1d1face00d Update Face Filter
- Remove old face filter
  - plugins.extract.pipeline: Expose plugins directly
  - Change `is_aligned` from plugin level to ExtractMedia level
  - Allow extract pipeline to take faceswap aligned images
  - Add ability for recognition plugins to accept aligned faces as input
  - Add face filter to recognition plugin
  - Move extractor pipeline IO ops to own class
2022-10-31 18:25:32 +00:00
torzdf 403f981e0f bugfix: Catch learn-mask error when no mask type selected 2022-10-12 12:03:59 +01:00
torzdf f1e3339fbe Add vggface2 to extraction pipeline 2022-10-10 13:09:02 +01:00
torzdf 1f31af233e bugfix: Aligner size filter to int64 2022-10-04 12:13:26 +01:00
torzdf 765e385177 Extract: Typing and standardization 2022-09-30 18:36:35 +01:00
torzdf 376c419498 Bugfix: Manual tool. Explicitly disable aligner filters 2022-09-24 01:09:41 +01:00
torzdf fba2e6e851 Bugfix: Alignments tool sorting error in 'from-faces' 2022-09-23 19:48:48 +01:00
torzdf 8776e21629 Extract: Add roll filter to aligner filters 2022-09-23 14:25:21 +01:00
torzdf 50d23792ca Aligner filter - set sensible defaults 2022-09-21 01:04:33 +01:00
torzdf f5fa5b4337 Extract filter: Allow saving of filtered images 2022-09-19 00:13:41 +01:00
torzdf 5ba07eaa58 bugfix: Extract - don't put empty batches through aligner 2022-09-18 20:44:01 +01:00
torzdf a8f22cc019 Extract updates:
- Default CPU detector to MTCNN
  - add basic Aligner false positive filters
  - Typing: align + plugins
  - Use specific AlignerBatch class for alignment
  -
2022-09-18 19:44:41 +01:00
torzdf 2d312a9db2 Minor updates and fixups
- Mask Tool - Typing + BiSeNet mask update fix
  - Alignments Tool - Auto search for alignments file
2022-09-14 19:14:03 +01:00
torzdf 98d01760e4 Overhaul sort:
- Standardize image data reading and writing
  - Optimize loading (just one pass required)
  - Make all sort groups binnable (to greater or lesser results)
  - Add sort by pitch
  - Deprecate multiple options
  - linting, docs + locales
2022-09-13 13:08:59 +01:00
torzdf 7da2cc3dd2 Training - Use custom preview pop-out 2022-09-07 11:49:52 +01:00
torzdf ae7793e876 bugfix: convert - Process final items on truncated batch 2022-09-02 12:42:52 +01:00
torzdf 477e3e2013 bugfix: Train: Toggle mask when 'learn_mask' is selected 2022-09-02 11:50:58 +01:00
torzdf 13cfb3f39e extract: Add batch processing mode 2022-08-31 19:48:47 +01:00
torzdf 97da6250ff typing fix 2022-08-30 12:39:59 +01:00
torzdf de1068b417 Bugfix: Preview - show predictions at correct size 2022-08-30 12:39:39 +01:00
torzdf f3b88d5626 bugfix: Get correct output size for learn mask 2022-08-29 01:04:40 +01:00
torzdf 1022651eb8 Bugfix: convert - Gif Writer
- Fix non-launch error on Gif Writer
  - convert plugins - linting
  - convert/fs_media/preview/queue_manager - typing
  - Change convert items from dict to Dataclass
2022-08-26 23:56:03 +01:00
torzdf 326110f09d bugfix - timelapse image loader
multithreading.py - typing + docs
2022-08-25 12:38:05 +01:00
torzdf 0837273d46 bugfix: CV2 error on previews 2022-08-22 21:31:51 +01:00
torzdf 2beceffad9 Data Augmentation update (#1263)
- lib.detected_face
    - Subclass Masks for Landmark based masks
    - Add training mask propery + methods to DetectedFace
  - lib.training_training
    - subclass TrainingDataGenerator for training and preview data
    - Split cache into own module
    - Reduce thread count to 1 to prevent image corruption + data re-use
    - Process on largest model input/output size rather than stored image size
    - Size and crop masks during caching stage
    - Implement ring buffer for data flow
    - Fix preview reload bug
  - augmentation
    - typing
    - switch color aug order
    - better initialization
    - Fix warp + landmark warp to correctly apply at different image scales
    - Slightly improved warp caching
    - Don't store whether image is_preview. Handle all data as training images implicitly
  - plugins.trainer: Typing and fixes to work with trainingdata refactor
2022-08-21 18:59:45 +01:00
torzdf 5e73437be4 lib.align updates:
- alignments.py
    - Add typed dicts for imported alignments
    - Explicitly check for presence of thumb value in alignments dict
    - linting
  - detected_face.py
    - Typing
    - Linting
    - Legacy support for pre-aligned face
    - Update dependencies to new property names
2022-08-19 12:36:08 +01:00
torzdf 3295089737 lib.detected_face.Mask
- Add source + target offset and coverage to set_sub_crop method
2022-08-19 10:57:40 +01:00
torzdf c0d0f04b70 bufix: mask plugin: Fix error when no mask is selected 2022-08-18 19:35:50 +01:00
torzdf 26dde3c19a Add CPU option to BiSeNet and MTCNN
- Add CPU option to KSession
  - MTCNN optimizations
  - Update docs for bisenet + mtcnn
2022-08-08 14:29:32 +01:00
torzdf 629c02a61e Add "custom" mask 2022-08-05 14:09:05 +01:00
torzdf 26e26c6288 Update TF to 2.9
- Update TF to 2.7 to 2.9
    - Bump dependencies
    - Remove decode from pynvml calls
    - force keras predict functions to non-verbose
    - update tests
    - update Tensorboard logging
    - Update docs
2022-08-02 17:12:26 +01:00