Final rename for everything

.flake8

@@ -1,5 +1,5 @@
 [flake8]
 select = E22, E23, E24, E27, E3, E4, E7, F, I1, I2
 plugins = flake8-import-order
-application_import_names = embedding_converter, face_swapper
+application_import_names = crossface, hyperswap
 import-order-style = pycharm

.github/workflows/ci.yml

@@ -15,8 +15,8 @@ jobs:
   - run: pip install flake8
   - run: pip install flake8-import-order
   - run: pip install mypy
-  - run: flake8 embedding_converter face_swapper
+  - run: flake8 crossface hyperswap
-  - run: mypy embedding_converter face_swapper
+  - run: mypy crossface hyperswap
  test:
   runs-on: ubuntu-latest
   steps:

crossface/README.md

@@ -1,7 +1,7 @@
-Embedding Converter
+CrossFace
-===================
+=========
 
-> Convert face embeddings between various models.
+> Seamless transform face embeddings across embedder models.
 
 ![License](https://img.shields.io/badge/license-OpenRAIL--MS-green)
 
@@ -9,7 +9,7 @@ Embedding Converter
 Preview
 -------
 
-![Preview](https://raw.githubusercontent.com/facefusion/facefusion-labs/master/.github/previews/embedding_converter.png?sanitize=true)
+![Preview](https://raw.githubusercontent.com/facefusion/facefusion-labs/master/.github/previews/crossface.png?sanitize=true)
 
 
 Installation
@@ -23,7 +23,7 @@ pip install -r requirements.txt
 Setup
 -----
 
-This `config.ini` utilizes the MegaFace dataset to train the Embedding Converter for SimSwap.
+This `config.ini` utilizes the MegaFace dataset to train the CrossFace model for SimSwap.
 
 ```
 [training.dataset]
@@ -50,13 +50,13 @@ max_epochs = 4096
 strategy = auto
 precision = 16-mixed
 logger_path = .logs
-logger_name = arcface_converter_simswap
+logger_name = crossface_simswap
 ```
 
 ```
 [training.output]
 directory_path = .outputs
-file_pattern = arcface_converter_simswap_{epoch}_{step}
+file_pattern = crossface_simswap_{epoch}_{step}
 resume_path = .outputs/last.ckpt
 ```
 
@@ -64,7 +64,7 @@ resume_path = .outputs/last.ckpt
 [exporting]
 directory_path = .exports
 source_path = .outputs/last.ckpt
-target_path = .exports/arcface_converter_simswap.onnx
+target_path = .exports/crossface_simswap.onnx
 ir_version = 10
 opset_version = 15
 ```
@@ -73,7 +73,7 @@ opset_version = 15
 Training
 --------
 
-Train the Embedding Converter model.
+Train the model.
 
 ```
 python train.py

crossface/src/exporting.py

@@ -3,7 +3,7 @@ from configparser import ConfigParser
 
 import torch
 
-from .training import EmbeddingConverterTrainer
+from .training import CrossFaceTrainer
 
 CONFIG_PARSER = ConfigParser()
 CONFIG_PARSER.read('config.ini')
@@ -17,7 +17,7 @@ def export() -> None:
 	config_opset_version = CONFIG_PARSER.getint('exporting', 'opset_version')
 
 	os.makedirs(config_directory_path, exist_ok = True)
-	model = EmbeddingConverterTrainer.load_from_checkpoint(config_source_path, map_location = 'cpu').eval()
+	model = CrossFaceTrainer.load_from_checkpoint(config_source_path, map_location ='cpu').eval()
 	model.ir_version = torch.tensor(config_ir_version)
 	input_tensor = torch.randn(1, 512)
 	torch.onnx.export(model, input_tensor, config_target_path, input_names = [ 'input' ], output_names = [ 'output' ], opset_version = config_opset_version)

crossface/src/models/crossface.py

@@ -2,7 +2,7 @@ import torch
 from torch import Tensor, nn
 
 
-class EmbeddingConverter(nn.Module):
+class CrossFace(nn.Module):
 	def __init__(self) -> None:
 		super().__init__()
 		self.layers = self.create_layers()

crossface/src/training.py

@@ -11,26 +11,26 @@ from torch.utils.data import Dataset, random_split
 from torchdata.stateful_dataloader import StatefulDataLoader
 
 from .dataset import StaticDataset
-from .models.embedding_converter import EmbeddingConverter
+from .models.crossface import CrossFace
 from .types import Batch, Embedding, OptimizerSet
 
 CONFIG_PARSER = ConfigParser()
 CONFIG_PARSER.read('config.ini')
 
 
-class EmbeddingConverterTrainer(LightningModule):
+class CrossFaceTrainer(LightningModule):
 	def __init__(self, config_parser : ConfigParser) -> None:
 		super().__init__()
 		self.config_source_path = config_parser.get('training.model', 'source_path')
 		self.config_target_path = config_parser.get('training.model', 'target_path')
 		self.config_learning_rate = config_parser.getfloat('training.trainer', 'learning_rate')
-		self.embedding_converter = EmbeddingConverter()
+		self.crossface = CrossFace()
 		self.source_embedder = torch.jit.load(self.config_source_path, map_location = 'cpu').eval()
 		self.target_embedder = torch.jit.load(self.config_target_path, map_location = 'cpu').eval()
 		self.mse_loss = nn.MSELoss()
 
 	def forward(self, source_embedding : Embedding) -> Embedding:
-		return self.embedding_converter(source_embedding)
+		return self.crossface(source_embedding)
 
 	def training_step(self, batch : Batch, batch_index : int) -> Tensor:
 		with torch.no_grad():
@@ -125,10 +125,10 @@ def train() -> None:
 
 	dataset = StaticDataset(CONFIG_PARSER)
 	training_loader, validation_loader = create_loaders(dataset)
-	embedding_converter_trainer = EmbeddingConverterTrainer(CONFIG_PARSER)
+	crossface_trainer = CrossFaceTrainer(CONFIG_PARSER)
 	trainer = create_trainer()
 
 	if os.path.exists(config_resume_path):
-		trainer.fit(embedding_converter_trainer, training_loader, validation_loader, ckpt_path = config_resume_path)
+		trainer.fit(crossface_trainer, training_loader, validation_loader, ckpt_path = config_resume_path)
 	else:
-		trainer.fit(embedding_converter_trainer, training_loader, validation_loader)
+		trainer.fit(crossface_trainer, training_loader, validation_loader)

hyperswap/README.md

@@ -1,7 +1,7 @@
-Face Swapper
+HyperSwap
-============
+=========
 
-> Face shape and occlusion aware identity transfer.
+> Hyper accurate face swapping for everyone.
 
 ![License](https://img.shields.io/badge/license-ResearchRAIL--MS-red)
 
@@ -23,12 +23,12 @@ pip install -r requirements.txt
 Setup
 -----
 
-This `config.ini` utilizes the VGGFace2 dataset to train the Face Swapper model.
+This `config.ini` utilizes the VGGFace2 dataset to train the HyperSwap model.
 
 ```
 [training.dataset]
 file_pattern = .datasets/vggface2/**/*.jpg
-warp_template = vgg_face_hq_to_arcface_128_v2
+warp_template = vggfacehq_256_to_arcface_128_v2
 transform_size = 256
 batch_mode = equal
 batch_ratio = 0.2
@@ -92,14 +92,14 @@ max_epochs = 50
 strategy = auto
 precision = 16-mixed
 logger_path = .logs
-logger_name = face_swapper
+logger_name = hyperswap
 preview_frequency = 100
 ```
 
 ```
 [training.output]
 directory_path = .outputs
-file_pattern = face_swapper_{epoch}_{step}
+file_pattern = hyperswap_{epoch}_{step}
 resume_path = .outputs/last.ckpt
 ```
 
@@ -127,7 +127,7 @@ output_path = .outputs/output.jpg
 Training
 --------
 
-Train the Face Swapper model.
+Train the model.
 
 ```
 python train.py

hyperswap/src/exporting.py

@@ -5,7 +5,7 @@ from typing import Tuple
 import torch
 from torch import Tensor, nn
 
-from .training import FaceSwapperTrainer
+from .training import HyperSwapTrainer
 from .types import Embedding, Mask, Module
 
 CONFIG_PARSER = ConfigParser()
@@ -36,7 +36,7 @@ def export() -> None:
 	config_precision = CONFIG_PARSER.get('exporting', 'precision')
 
 	os.makedirs(config_directory_path, exist_ok = True)
-	model = FaceSwapperTrainer.load_from_checkpoint(config_source_path, config_parser = CONFIG_PARSER, map_location = 'cpu').eval()
+	model = HyperSwapTrainer.load_from_checkpoint(config_source_path, config_parser = CONFIG_PARSER, map_location ='cpu').eval()
 
 	if config_precision == 'half':
 		model = HalfPrecision(model).eval()

hyperswap/src/helper.py

@@ -10,12 +10,12 @@ WARP_TEMPLATE_SET : WarpTemplateSet =\
 		[ 8.75000016e-01, -1.07193451e-08, 3.80446920e-10 ],
 		[ 1.07193451e-08, 8.75000016e-01, -1.25000007e-01 ]
 	]),
-	'ffhq_to_arcface_128_v2': torch.tensor(
+	'ffhq_512_to_arcface_128_v2': torch.tensor(
 	[
 		[ 8.50048894e-01, -1.29486822e-04, 1.90956388e-03 ],
 		[ 1.29486822e-04, 8.50048894e-01, 9.56254653e-02 ]
 	]),
-	'vgg_face_hq_to_arcface_128_v2': torch.tensor(
+	'vggfacehq_256_to_arcface_128_v2': torch.tensor(
 	[
 		[ 1.01305414, -0.00140513, -0.00585911 ],
 		[ 0.00140513, 1.01305414, 0.11169602 ]

hyperswap/src/inferencing.py

@@ -4,7 +4,7 @@ import torch
 from torchvision import io
 
 from .helper import calc_embedding
-from .training import FaceSwapperTrainer
+from .training import HyperSwapTrainer
 
 CONFIG_PARSER = configparser.ConfigParser()
 CONFIG_PARSER.read('config.ini')
@@ -17,7 +17,7 @@ def infer() -> None:
 	config_target_path = CONFIG_PARSER.get('inferencing', 'target_path')
 	config_output_path = CONFIG_PARSER.get('inferencing', 'output_path')
 
-	generator = FaceSwapperTrainer.load_from_checkpoint(config_generator_path, config_parser = CONFIG_PARSER, map_location = 'cpu').eval()
+	generator = HyperSwapTrainer.load_from_checkpoint(config_generator_path, config_parser = CONFIG_PARSER, map_location ='cpu').eval()
 	embedder = torch.jit.load(config_embedder_path, map_location = 'cpu').eval()
 
 	source_tensor = io.read_image(config_source_path)

hyperswap/src/training.py

@@ -25,7 +25,7 @@ CONFIG_PARSER = ConfigParser()
 CONFIG_PARSER.read('config.ini')
 
 
-class FaceSwapperTrainer(LightningModule):
+class HyperSwapTrainer(LightningModule):
 	def __init__(self, config_parser : ConfigParser) -> None:
 		super().__init__()
 		self.config_generator_embedder_path = config_parser.get('training.model', 'generator_embedder_path')
@@ -239,10 +239,10 @@ def train() -> None:
 
 	dataset = ConcatDataset(prepare_datasets(CONFIG_PARSER))
 	training_loader, validation_loader = create_loaders(dataset)
-	face_swapper_trainer = FaceSwapperTrainer(CONFIG_PARSER)
+	hyperswap_trainer = HyperSwapTrainer(CONFIG_PARSER)
 	trainer = create_trainer()
 
 	if os.path.isfile(config_resume_path):
-		trainer.fit(face_swapper_trainer, training_loader, validation_loader, ckpt_path = config_resume_path)
+		trainer.fit(hyperswap_trainer, training_loader, validation_loader, ckpt_path = config_resume_path)
 	else:
-		trainer.fit(face_swapper_trainer, training_loader, validation_loader)
+		trainer.fit(hyperswap_trainer, training_loader, validation_loader)

hyperswap/src/types.py

@@ -20,5 +20,5 @@ FaceMaskerModule : TypeAlias = Module
 
 OptimizerSet : TypeAlias = Any
 
-WarpTemplate = Literal['arcface_128_v2_to_arcface_112_v2', 'ffhq_to_arcface_128_v2', 'vgg_face_hq_to_arcface_128_v2']
+WarpTemplate = Literal['arcface_128_v2_to_arcface_112_v2', 'ffhq_512_to_arcface_128_v2', 'vggfacehq_256_to_arcface_128_v2']
 WarpTemplateSet : TypeAlias = Dict[WarpTemplate, Tensor]

hyperswap/tests/test_networks.py

@@ -3,9 +3,9 @@ from configparser import ConfigParser
 import pytest
 import torch
 
-from face_swapper.src.networks.aad import AAD
+from hyperswap.src.networks.aad import AAD
-from face_swapper.src.networks.masknet import MaskNet
+from hyperswap.src.networks.masknet import MaskNet
-from face_swapper.src.networks.unet import UNet
+from hyperswap.src.networks.unet import UNet
 
 
 @pytest.mark.parametrize('output_size', [ 128, 256, 512 ])