Dataset Viewer
Auto-converted to Parquet Duplicate
type
string
coordinates
list
Polygon
[[[-91.51307624682767,40.17853276735267],[-91.51280694860233,40.17588464132158],[-91.51271713366296,(...TRUNCATED)

YAML Metadata Warning:The task_categories "image-regression" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

rs-embed I-GUIDE tutorial dataset — Illinois maize yield

Precomputed foundation-model embeddings for the rs-embed tutorial notebook (examples/demo_iguide.ipynb): 991 SPAM 2020 maize-yield points across Illinois, embedded by 14 remote-sensing foundation models from a Sentinel-2 (and MODIS for SatVision) summer 2019 median composite, then adaptive-average-pooled to at most 8×8 spatial resolution.

Files

File Contents
labels.csv One row per point: grid_code, lon, lat, maize_area_ha, label (SPAM 2020 maize yield, t/ha), split (fixed train/test, 75/25, seed 42)
features/<model>.npz Key X: float16 array (991, C, t, t), t = min(8, H, W). Row i aligns with row i of labels.csv. Rows where the model failed are NaN.
thumbnails.npz Key rgb: uint8 (991, 128, 128, 3) Sentinel-2 true-color previews (2–98% percentile stretch)
illinois_boundary.json GeoJSON geometry of the Illinois state boundary (TIGER/2018)
showcase/p00144.npz One location at full resolution for segmentation demos: keys emb__<model> (float16 (C, H, W) native grids) and rgb (uint8 true-color patch)

Models: agrifm, clay, dofa, fomo, gse, olmoearth, remoteclip, satmaepp, satvision, terrafm, terramind, tessera, thor, wildsat.

How it was made

Embeddings were exported with a single rs_embed.export_batch(...) call (grid output, GEE backend, temporal range 2019-06-01 → 2019-08-31, 2048 m point buffers), then pooled with examples/iguide/prepare_hf_dataset.py. Pooling uses the same binning as torch.nn.functional.adaptive_avg_pool2d.

Sources

  • Labels: SPAM 2020 v2r0 global harvested area / yield (CC BY 4.0)
  • Imagery: Copernicus Sentinel-2 L2A and MODIS MOD021KM via Google Earth Engine
  • Boundary: US Census TIGER/2018 States
Downloads last month
76