Colab notebook is available now! You can experience FaceChain directly with . (August 15th, 2023 UTC) FaceChain is a deep-learning toolchain for generating your Digital-Twin. With a minimum of 1 portrait-photo, you can create a Digital-Twin of your own and start generating personal portraits in different settings (multiple styles now supported!). You may train your Digital-Twin model and generate photos via FaceChain's Python scripts, or via the familiar Gradio interface. You can also experience FaceChain directly with our ModelScope Studio. Add robust face lora training module, enhance the performance of one pic training & style-lora blending. (August 27th, 2023 UTC)

The ModelScope notebook has a free tier that allows you to run the FaceChain application, refer to ModelScope Notebook In addition to ModelScope notebook and ECS, I would suggest that we add that user may also start DSW instance with the option of ModelScope (GPU) image, to create a ready-to-use environment. FaceChain is a deep-learning toolchain for generating your Digital-Twin. With a minimum of 1 portrait-photo, you can create a Digital-Twin of your own and start generating personal portraits in different settings (multiple styles now supported!). You may train your Digital-Twin model and generate photos via FaceChain's Python scripts, or via the familiar Gradio interface, or via sd webui. Note: FaceChain currently assume single-GPU, if your environment has multiple GPU, please use the following instead: # CUDA_VISIBLE_DEVICES=0 python3 app.py # Step6 Support super resolution🔥🔥🔥, provide multiple resolution choice (512 512, 768768, 1024 1024, 20482048). (November 13th, 2023 UTC) Step5 clone facechain from github GIT_LFS_SKIP_SMUDGE = 1 git clone https://github.com/modelscope/facechain.git --depth 1 cd facechain

Help

Face recognition model RTS： https://modelscope.cn/models/damo/cv_ir_face-recognition-ood_rts More Information Face quality assessment FQA： https://modelscope.cn/models/damo/cv_manual_face-quality-assessment_fqa FaceChain supports direct training and inference in the python environment. Run the following command in the cloned folder to start training: PYTHONPATH =. sh train_lora.sh "ly261666/cv_portrait_model" "v2.0" "film/film" "./imgs" "./processed" "./output"

When inferring, please edit the code in run_inference.py: # Fill in the folder of the images after preprocessing above, it should be the same as during training processed_dir = './processed' # The number of images to generate in inference num_generate = 5 # The stable diffusion base model used in training, no need to be changed base_model = 'ly261666/cv_portrait_model' # The version number of this base model, no need to be changed revision = 'v2.0' # This base model may contains multiple subdirectories of different styles, currently we use film/film, no need to be changed base_model_sub_dir = 'film/film' # The folder where the model weights stored after training, it must be the same as during training train_output_dir = './output' # Specify a folder to save the generated images, this parameter can be modified as needed output_dir = './generated' Face attribute recognition model FairFace： https://modelscope.cn/models/damo/cv_resnet34_face-attribute-recognition_fairface imgs: This parameter needs to be replaced with the actual value. It means a local file directory that contains the original photos used for training and generationStep1: 我的notebook -> PAI-DSW -> GPU环境 # Step2: Open the Terminal，clone FaceChain from github: GIT_LFS_SKIP_SMUDGE = 1 git clone https://github.com/modelscope/facechain.git --depth 1 # Step3: Entry the Notebook cell: GIT_LFS_SKIP_SMUDGE = 1 git clone https://github.com/modelscope/facechain.git --depth 1 cd facechain You can find the generated personal digital image photos in the output_dir. Algorithm Introduction Architectural Overview

Face detection model DamoFD： https://modelscope.cn/models/damo/cv_ddsar_face-detection_iclr23-damofd

FaceChain has been selected in the BenchCouncil Open100 (2022-2023) annual ranking. (November 8th, 2023 UTC) Use the conda virtual environment, and refer to Anaconda to manage your dependencies. After installation, execute the following commands: High performance inpainting for single & double person, Simplify User Interface. (September 09th, 2023 UTC) Input: User-uploaded images in the training phase, preset input prompt words for generating personal portraits Use depth control, default False, only effective when using pose control use_depth_control = False # Use pose control, default False use_pose_model = False # The path of the image for pose control, only effective when using pose control pose_image = 'poses/man/pose1.png' # Fill in the folder of the images after preprocessing above, it should be the same as during training processed_dir = './processed' # The number of images to generate in inference num_generate = 5 # The stable diffusion base model used in training, no need to be changed base_model = 'ly261666/cv_portrait_model' # The version number of this base model, no need to be changed revision = 'v2.0' # This base model may contains multiple subdirectories of different styles, currently we use film/film, no need to be changed base_model_sub_dir = 'film/film' # The folder where the model weights stored after training, it must be the same as during training train_output_dir = './output' # Specify a folder to save the generated images, this parameter can be modified as needed output_dir = './generated' # Use Chinese style model, default False use_style = False

Human parsing model M2FP： https://modelscope.cn/models/damo/cv_resnet101_image-multiple-human-parsingHuggingFace Space is available now! You can experience FaceChain directly with 🤗 (August 25th, 2023 UTC)