config.json

{
  "model_type": "vit",
  "task": "normal_vs_pvd",
  "pipeline_tag": "image-classification",
  "num_labels": 2,
  "id2label": {
    "0": "Normal",
    "1": "Posterior Vitreous Detachment"
  },
  "label2id": {
    "Normal": 0,
    "Posterior Vitreous Detachment": 1
  },
  "input_modality": "3D ocular ultrasound video",
  "input_shape": [
    1,
    96,
    128,
    128
  ],
  "input_format": "video (mp4), normalized by /255.0",
  "architecture": {
    "in_channels": 1,
    "num_classes": 1,
    "img_size": [
      96,
      128,
      128
    ],
    "patch_size": 7,
    "hidden_size": 768,
    "num_layers": 4,
    "num_heads": 4
  },
  "pooling": "avg"
}