config.json

{
  "model_name": "RemoteCLIP",
  "model_type": "remoteclip",
  "architectures": [
    "RemoteCLIP"
  ],
  "framework": "PyTorch",
  "domain": "earth-observation",
  "task": "remote-sensing-image-text-retrieval",
  "implementation": {
    "entry_point": "model/remoteclip.py",
    "scope": "CLIP-compatible vision and text dual encoders with multi-positive symmetric contrastive learning"
  },
  "architecture": {
    "family": "vision transformer and causal text transformer dual encoder",
    "image_shape": [
      3,
      224,
      224
    ],
    "patch_size": 32,
    "context_length": 77,
    "vocabulary_size": 49408,
    "embed_dim": 64,
    "vision_width": 64,
    "vision_layers": 2,
    "vision_heads": 4,
    "text_width": 64,
    "text_layers": 2,
    "text_heads": 4
  },
  "data": {
    "datasets": [
      "RSITMD",
      "RSICD",
      "UCM-Captions"
    ],
    "protocol": "remoteclip_rgb_bpe_pairs_v1",
    "input_format": "image_text_npz",
    "train_samples": 8,
    "test_samples": 4
  },
  "configuration_sources": [
    "configuration.json",
    "conf/config.yaml",
    "model/remoteclip.py"
  ]
}