| { |
| "model_name": "TerraMind", |
| "model_type": "terramind", |
| "architectures": ["TerraMind", "Transformer"], |
| "framework": "PyTorch", |
| "domain": "earth-observation", |
| "task": "multimodal-any-to-any-generation", |
| "implementation": { |
| "entry_point": "model/terramind.py", |
| "scope": "dual-scale multimodal masked-token engineering reproduction with conditional any-to-any generation", |
| "train_script": "scripts/train.py", |
| "inference_script": "scripts/inference.py", |
| "evaluation_script": "scripts/result.py", |
| "synthetic_data_script": "scripts/fake_data.py" |
| }, |
| "architecture": { |
| "family": "dual-scale multimodal transformer encoder-decoder", |
| "source_size": 264, |
| "input_size": 224, |
| "patch_size": 16, |
| "patch_grid": [14, 14], |
| "dim": 96, |
| "encoder_depth": 2, |
| "decoder_depth": 2, |
| "heads": 4, |
| "engineering_vocab_size": 256, |
| "visible_fraction": 0.35, |
| "pixel_modalities": {"s2l2a": 12, "s2l1c": 13, "s2rgb": 3, "s1grd": 2, "s1rtc": 2, "dem": 1}, |
| "token_modalities": ["s2l2a", "s1grd", "s1rtc", "dem", "lulc", "ndvi", "coords", "caption"] |
| }, |
| "data": { |
| "datasets": ["TerraMesh"], |
| "protocol": "terramind_synthetic_engineering_v2", |
| "format": "NPZ", |
| "train_file": "data/train.npz", |
| "test_file": "data/test.npz", |
| "source_spatial_shape": [264, 264], |
| "model_spatial_shape": [224, 224], |
| "required_metadata": ["format_version", "data_source", "coords", "caption", "token_map_lulc"] |
| }, |
| "configuration_sources": ["conf/config.yaml", "model/terramind.py", "scripts/fake_data.py", "scripts/train.py", "scripts/inference.py", "scripts/result.py"] |
| } |
|
|
