| { |
| "model_name": "SatMAE++", |
| "model_type": "satmae_pp", |
| "architectures": [ |
| "SatMAEPP" |
| ], |
| "framework": "PyTorch", |
| "domain": "earth-observation", |
| "task": "remote-sensing-representation-learning", |
| "implementation": { |
| "entry_point": "model/satmaepp.py", |
| "scope": "masked autoencoding with convolutional native-scale reconstruction for RGB or grouped multispectral satellite imagery", |
| "train_script": "scripts/train.py", |
| "inference_script": "scripts/inference.py", |
| "evaluation_script": "scripts/result.py", |
| "synthetic_data_script": "scripts/fake_data.py" |
| }, |
| "architecture": { |
| "family": "multiscale masked autoencoder with convolutional upsampling", |
| "runtime_profile": "smoke", |
| "mode": "rgb", |
| "image_size": 32, |
| "patch_size": 4, |
| "in_channels": 3, |
| "embed_dim": 64, |
| "encoder_depth": 2, |
| "encoder_heads": 4, |
| "decoder_dim": 32, |
| "decoder_depth": 1, |
| "decoder_heads": 4, |
| "mask_ratio": 0.75, |
| "scales": [1, 2], |
| "spectral_groups": null, |
| "spatial_mask": false, |
| "norm_pix_loss": false, |
| "proj_ratio": 4 |
| }, |
| "data": { |
| "datasets": [ |
| "FMoW-RGB", |
| "FMoW-Sentinel", |
| "EuroSAT", |
| "UCMerced", |
| "RESISC-45" |
| ], |
| "protocol": "fmow_rgb_satmaepp", |
| "format": "NPZ", |
| "input_key": "images", |
| "input_shape": ["N", 3, 32, 32], |
| "native_target_keys": ["images_2x"], |
| "native_target_shapes": [["N", 3, 64, 64]], |
| "label_key": "labels", |
| "label_shape": ["N"], |
| "num_classes": 10, |
| "default_train_file": "data/train.npz", |
| "default_test_file": "data/test.npz", |
| "required_metadata": ["protocol", "data_source"] |
| }, |
| "configuration_sources": [ |
| "conf/config.yaml", |
| "model/satmaepp.py", |
| "scripts/fake_data.py", |
| "scripts/train.py", |
| "scripts/inference.py", |
| "scripts/result.py" |
| ] |
| } |
|
|
