Usage: This GAN is used as part of the medigan library. This GANs metadata is therefore stored in and retrieved from medigan-model's config file. medigan is an open-source Python library on Github that allows developers and researchers to easily add synthetic imaging data into their model training pipelines. medigan is documented here and can be used via pip install: pip install medigan To run this model in medigan, use the following commands. # import medigan and initialize generators from medigan import Generators generators = Generators() # Generate 10 images generators.generate(model_id="00003_CYCLEGAN_MMG_DENSITY_FULL",num_samples=10) Description: A cycle generative adversarial network (CycleGAN) that generates mammograms with high breast density from an original mammogram e.g. with low-breast density. The CycleGAN was trained using normal (without pathologies) digital mammograms from BCDR dataset (Lopez, M. G., et al. 2012). The uploaded ZIP file contains the files CycleGAN_high_density.pth (model weights), __init__.py (image generation method and utils), a license, and the GAN model architecture (in pytorch) below the /src folder. Note: The images in the CycleGAN_high_density/images are synthetically generated low breast density mammograms. These images were generated with a high-to-low breast density mammogram translating cycleGAN (also trained on BCDR) with similar architecture to the present one. These synthetic mammogram images were uploaded instead of original images to avoid any conflicts in regard to copyright and intellectual property. The synthetic images are example images that medigan users may translate to examine how the present cycleGAN works. In particular, these synthetic images will be randomly translated if the medigan users do not provide their own input images to the present cycleGAN. Metadata: The source of the metadata displayed below is the global.json file in the medigan-models Github repository. { "00003_CYCLEGAN_MMG_DENSITY_FULL": { "execution": { "package_name": "CycleGAN_high_density", "package_link": "https://zenodo.org/record/5547264/files/CycleGAN_high_density.zip?download=1", "model_name": "CycleGAN_high_density", "extension": ".pth", "image_size": [ 1332, 800 ], "dependencies": [ "numpy", "Path", "Union", "pyyaml", "opencv-contrib-python-headless", "torch", "torchvision", "dominate", "visdom", "Pillow" ], "generate_method": { "name": "generate_GAN_images", "args": { "base": [ "model_file", "output_path", "save_images", "num_samples" ], "custom": { "translate_all_images": false, "input_path": "00003_CYCLEGAN_MMG_DENSITY_FULL/CycleGAN_high_density/images", "image_size": [1332, 800], "gpu_id": 0 } } } }, "selection": { "performance": {}, "use_cases": [ "classification", "detection", "domain-translation" ], "organ": [ "breast", "breasts", "chest" ], "modality": [ "MMG", "Mammography", "Mammogram", "full-field digital", "full-field digital MMG", "full-field MMG", "full-field Mammography", "digital Mammography", "digital MMG", "x-ray mammography" ], "vendors": [], "centres": [], "function": [ "image to image", "image generation", "data augmentation" ], "condition": [], "dataset": [ "BCDR" ], "augmentations": [ "resize" ], "generates": [ "full images", "mammograms", "full-field digital mammograms" ], "height": 1332, "width": 800, "depth": null, "type": "CycleGAN", "license": "BSD", "dataset_type": "public", "privacy_preservation": null, "tags": [ "Mammogram", "Mammography", "Digital Mammography", "Full field Mammography", "Full-field Mammography", "CycleGANs", "CycleGAN", "Density", "Breast Density", "High Density", "Low Density", "ACR" ], "year": "2021" }, "description": { "title": "CycleGAN Model for Low-to-High Brest Density Mammograms Translation (Trained on BCDR)", "provided_date": "12th Sep 2021", "trained_date": "Sep 2021", "provided_after_epoch": 100, "version": "0.0.1", "publication": null, "doi": [ "10.5281/zenodo.5547263" ], "comment": "A cycle generative adversarial network (CycleGAN) that generates mammograms with high breast density from an original mammogram e.g. with low-breast density. The CycleGAN was trained using normal (without pathologies) digital mammograms from BCDR dataset (Lopez, M. G., et al. 2012). The uploaded ZIP file contains the files CycleGAN_high_density.pth (model weights), __init__.py (image generation method and utils) and the GAN model architecture (in pytorch) below the /src folder." } } }