Model Integration Request for medigan: 00013_C-DCGAN_MMG_MASSES

[RichardObi]

Creator: unknown name
Affiliation: unknown affiliation
Stored in: https://sandbox.zenodo.org/record/1076278

Model Metadata:

  "00013_C-DCGAN_MMG_MASSES": {
     "execution": {
        "package_name": "C-DCGAN_MMG_MASSES_BCDR_MAL_BEN",
        "package_link": "https://sandbox.zenodo.org/record/1076278",
        "model_name": "1250",
        "extension": ".pt",
        "image_size": [
           128,
           128
        ],
        "dependencies": [
           "numpy",
           "torch",
           "opencv-contrib-python-headless"
        ],
        "generate_method": {
           "name": "generate",
           "args": {
              "base": [
                 "model_file",
                 "num_samples",
                 "output_path",
                 "save_images"
              ],
              "custom": {
                 "condition": null,
                 "z": null
              }
           },
           "input_latent_vector_size": 100
        }
     },
     "selection": {
        "performance": {
           "SSIM": null,
           "MSE": null,
           "NSME": null,
           "PSNR": null,
           "IS": null,
           "FID": null,
           "turing_test": null,
           "downstream_task": {
              "CLF": {}
           }
        },
        "use_cases": [
           "classification",
           "malignant versus benign classification"
        ],
        "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": [
           "noise to image",
           "image generation",
           "unconditional generation",
           "data augmentation"
        ],
        "condition": [],
        "dataset": [
           "CBIS-DDSM"
        ],
        "augmentations": [
           "horizontal flip",
           "vertical flip"
        ],
        "generates": [
           "mass",
           "masses",
           "mass roi",
           "mass ROI",
           "mass images",
           "mass region of interest",
           "nodule",
           "nodule",
           "nodule roi",
           "nodule ROI",
           "nodule images",
           "nodule region of interest"
        ],
        "height": 128,
        "width": 128,
        "depth": null,
        "type": "Conditional DCGAN",
        "license": "MIT",
        "dataset_type": "public",
        "privacy_preservation": null,
        "tags": [
           "Breast",
           "Mammogram",
           "Mammography",
           "Digital Mammography",
           "Full field Mammography",
           "Full-field Mammography",
           "128x128",
           "128 x 128",
           "MammoGANs",
           "Masses",
           "Nodules"
        ],
        "year": "2022"
     },
     "description": {
        "title": "Conditional DCGAN Model for Patch Generation of Mammogram Masses Conditioned on Biopsy Proven Malignancy Status (Trained on BCDR)",
        "provided_date": "June 2022",
        "trained_date": "June 2022",
        "provided_after_epoch": 1250,
        "version": "1.0.0",
        "publication": null,
        "doi": [],
        "comment": "A class-conditional deep convolutional generative adversarial network that generates mass patches of mammograms that are conditioned to either be benign (1) or malignant (0). Pixel dimensions are 128x128. The Cond-DCGAN was trained on MMG patches from the BCDR dataset (Lopez et al, 2012). The uploaded ZIP file contains the files 1250.pt (model weight), __init__.py (image generation method and utils), a requirements.txt, a LICENSE file, the MEDIGAN metadata, the used GAN training config file, a test.sh file to run the model, and two folders with a few generated images."
     }
  }
}