RichardObi / medigan

medigan - A Python Library of Pretrained Generative Models for Medical Image Synthesis
https://medigan.readthedocs.io/en/latest/
MIT License
115 stars 13 forks source link

Model Integration Request for medigan: 00019_PGGAN_CHEST_XRAY #32

Closed RichardObi closed 1 year ago

RichardObi commented 1 year ago

Creator: Eloy Garcia, Richard Osuala Affiliation: University of Barcelona Stored in: https://zenodo.org/record/6943804

Model Metadata:


   "00019_PGGAN_CHEST_XRAY": {
      "execution": {
         "package_name": "00019_PGGAN_NODE21",
         "package_link": "https://zenodo.org/record/6943804",
         "model_name": "model",
         "extension": ".pt",
         "image_size": [
            1024,
            1024,
            1
         ],
         "dependencies": [
            "numpy",
            "tqdm",
            "torch"
         ],
         "generate_method": {
            "name": "generate",
            "args": {
               "base": [
                  "model_file",
                  "num_samples",
                  "output_path",
                  "save_images"
               ],
               "custom": {
                  "image_size": 1024
               }
            },
            "real_input_latent_vector_size": 512,
            "input_latent_vector_size": 1024
         }
      },
      "selection": {
         "performance": {
            "SSIM": null,
            "MSE": null,
            "NSME": null,
            "PSNR": null,
            "IS": null,
            "FID": null,
            "turing_test": null,
            "downstream_task": null
         },
         "use_cases": [
            "classification",
            "detection"
         ],
         "organ": [
            "lung",
            "chest",
            "thorax"
         ],
         "modality": [
            "x-ray",
            "xray",
            "CXR"
         ],
         "vendors": [],
         "centres": [],
         "function": [
            "noise to image",
            "unconditional generation",
            "data augmentation"
         ],
         "condition": [],
         "dataset": [
            "Node21"
         ],
         "augmentations": [
            "horizontal flip",
            "vertical flip"
         ],
         "generates": [
            "chest xray",
            "CXR",
            "thoracic xray",
            "lung xray",
            "lung xray"
         ],
         "height": 128,
         "width": 128,
         "depth": 1,
         "type": "WGANGP",
         "license": "MIT",
         "dataset_type": "public",
         "privacy_preservation": null,
         "tags": [
            "Thoracic xray",
            "xray",
            "x-ray",
            "Thorax",
            "Lung",
            "Nodules",
            "Lung Cancer",
            "Lung Tumor"
         ],
         "year": "2022"
      },
      "description": {
         "title": "PGGAN Model for Generation of Chest XRAY (CXR) Images (Trained on Node21)",
         "provided_date": "June 2022",
         "trained_date": "June 2022",
         "provided_after_epoch": null,
         "version": null,
         "publication": null,
         "doi": [],
         "inputs": "image_size: default=1024, help=the size if height and width of the generated images.",
         "comment": "A unconditional Progressively-growing generative adversarial network (PGGAN) that generates chest xray (CXR) images with pixel dimensions 1024x1024. The PGGAN was trained on CXR images from the NODE21 dataset (Sogancioglu et al, 2021). The uploaded ZIP file contains the files model.pt (model weight), __init__.py (image generation method and utils), a requirements.txt, a LICENSE file, the MEDIGAN metadata.json file, the used GAN training config file, a test.sh file to run the model, and an /image folder with a few generated example images."
      }
   }
}