{"success":true,"data":{"author":"S. M. Hemant","bucketAccessType":"public","bucketName":"hem436/lunar-south-pole-trn-dataset","bucketPrefix":"","category":"Astronomy & Space","cloudProvider":"huggingface","createdAt":"2026-11-18T00:06:04.307Z","description":"1,380 path-traced orthographic renders of lunar south polar terrain for terrain relative navigation (TRN), map-relative localization and illumination-invariant feature matching. Six sites (lm7, site04, site11, site20, site42, sl2) built from NASA LRO LOLA south polar DEMs are split into 8 to 10 km tiles at 5 m per pixel in polar stereographic projection, and every tile is rendered under the same 60-point sun grid: 12 azimuths from 0 to 330 degrees times five grazing elevations from 0.11 to 4 degrees. Because the camera is fixed, a pixel is the same ground in all 60 renders, which gives exact correspondences across lighting. Images are 1600 to 2000 px 8-bit greyscale PNGs rendered in Blender Cycles with the JPL lunasynth generator, which stamps synthetic craters (2 to 20 m) and rocks (1.4 to 5 m); their positions ship as crater_field.parquet and rock_field.parquet per tile. index.csv lists every render with sun and camera parameters, and lunar_dataset.py provides site-level splits and illumination-pair loaders. About 1.8 GB. Licensed CC BY 4.0 by the author; cite the dataset, lunasynth and NASA LRO/LOLA.","displayTitle":"Lunar South Pole Synthetic Orthographic Renders for Terrain Relative Navigation (LOLA DEMs, 60-Sun Grid)","huggingFaceUrl":"https://huggingface.co/datasets/hem436/lunar-south-pole-trn-dataset","id":"cc8e0fa6-8d7c-5578-824f-e1123fd3c4c2","license":"CC-BY-4.0","licenseEvidenceUrl":"https://huggingface.co/datasets/hem436/lunar-south-pole-trn-dataset","objectCountEstimate":1442,"probeStatus":"ok","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"hem436/lunar-south-pole-trn-dataset\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"hem436/lunar-south-pole-trn-dataset\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=hem436, license:cc-by-4.0, size_categories:1K<n<10K, format:csv, downloads=121"],"format":"mixed","generatedAt":"2026-11-18T00:09:29.762Z","layout":{"example":"data/site04/lb/crater_field.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/site04/lb/crater_field.parquet","truncatedAt":512}],"sampleObjects":["data/lm7/lb/images/rgb_000.png","data/lm7/lb/images/rgb_001.png","data/lm7/lb/images/rgb_002.png","data/lm7/lb/images/rgb_003.png","data/lm7/lb/images/rgb_004.png"],"sidecars":[{"bytes":"Creative Commons Attribution 4.0 International (CC BY 4.0)\n\nCopyright (c) 2026 S.M.Hemant\n\nThis dataset is licensed under the Creative Commons Attribution 4.0\nInternational License. You are free to share and adapt the material for\nany purpose, including commercially, provided you give appropriate credit,\nlink to the license, and indicate if changes were made.\n\nThe complete and authoritative license text is published at:\n    https://creativecommons.org/licenses/by/4.0/legalcode\nA human-readable summary (not a substitute for the license) is at:\n    https://creativecommons.org/licenses/by/4.0/\nIn any conflict, the authoritative text at the URL above governs.\n\nATTRIBUTION NOTICE\nThe base elevation models are derived from NASA LRO LOLA lunar south pole\ndigital elevation models, which are US Government works in the public\ndomain and are not themselves covered by this license. Rendered imagery\nand synthetic crater/rock fields were generated with lunasynth, which is\nlicensed Apache-2.0 and is Copyright (c) Caltech/Jet Propulsion Laboratory.\n","encoding":"utf8","key":"LICENSE","name":"LICENSE","size":1049,"truncated":false},{"bytes":"---\nlicense: cc-by-4.0\ntask_categories:\n  - image-feature-extraction\n  - image-matching\ntags:\n  - lunar\n  - moon\n  - synthetic\n  - terrain-relative-navigation\n  - illumination-invariance\n  - remote-sensing\n  - blender\npretty_name: Lunar South Pole Synthetic Orthographic Renders for Terrain Relative Navigation\nsize_categories:\n  - n<1K\nconfigs:\n  - config_name: default\n    data_files:\n      - split: train\n        path: index.csv\n---\n\n# Lunar South Pole Synthetic Orthographic Renders for Terrain Relative Navigation\n\n1380 physically-based orthographic renders of lunar south polar terrain across\n6 sites, each tile rendered under the same 60-point sun grid.\nThe set is built for **illumination-invariant** work: every tile is rendered under\nthe same 60-point sun grid, so the identical terrain appears under\n60 different lighting conditions. That makes it directly usable for terrain\nrelative navigation (TRN), map-relative localisation, and feature-matching research\nwhere the hard part is that the same place looks completely different as the sun moves.\n\nGrazing illumination is the point. At the lunar south pole the sun never rises far\nabove the horizon, so shadows are long, high-contrast, and move dramatically with\nazimuth. Sun elevations here span 0.1146, 1, 2, 3, 4 degrees.\n\n## Dataset at a glance\n\n| | |\n|---|---|\n| Renders | 1380 |\n| Image size | 1600 x 1600 / 1680 x 1680 / 2000 x 2000 px, 8-bit greyscale PNG |\n| Ground sample distance | 5 m/px, identical across every site |\n| Footprint | 8 / 8.4 / 10 km square per image, depending on site |\n| Camera | Orthographic, nadir (pitch -90 deg), centred on the tile |\n| Sun elevations | 0.1146, 1, 2, 3, 4 degrees |\n| Sun azimuths | 0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, 330 degrees |\n| Exact angles | see `index.csv`; stored in radians upstream, so degrees carry sub-0.001 rounding |\n| Renderer | Blender Cycles, path traced, 4096 max samples (adaptive), denoised |\n| Total size | ~1.8 GB |\n\n## Sites\n\n| Site | Tiles | Renders | Image | Ground truth | Notes |\n|---|---|---|---|---|---|\n| `lm7` | 4 | 240 | 2000 px | none | 10 km tiles at 2000 px |\n| `site04` | 4 | 240 | 1680 px | craters + rocks | 16 x 16 km LOLA DEM, quartered |\n| `site11` | 4 | 240 | 1600 / 1680 px | rocks | connecting ridge; tile `lb` is 1680 px / 8.4 km, the other three 1600 px / 8 km |\n| `site20` | 4 | 240 | 1680 px | craters + rocks | 16 x 16 km LOLA DEM, quartered |\n| `site42` | 4 | 240 | 1680 px | craters + rocks | 20 x 20 km LOLA DEM, centre-cropped to 16 km then quartered |\n| `sl2` | 3 | 180 | 1600 px | craters + rocks | Gerlache top; only three quadrants exist, `rb` was never rendered |\n\nEvery site is lunar south polar, in a polar stereographic projection\n(`lat_0 = -90`) sampled at 5 m/px. Each source DEM is split into quadrants named\n`lt`, `rt`, `lb`, `rb` (left/right, top/bottom); the older campaigns used\n`tl`/`tr`/`bl`/`br` for the same quadrants and have been renamed to this one\nconvention. Site 42's source DEM is 20 x 20 km and was centre-cropped to 16 km\nbefore splitting, so its tiles match the other 8 km sites.\n\nThe sites were rendered in separate campaigns, so tile extent and image size\ndiffer between them (8 to 10 km, 1600 to 2000 px) even though the ground sample\ndistance is 5 m/px throughout. Use the per-row `width_px`, `footprint_km` and\n`ground_sample_distance_m` columns in `index.csv` rather than assuming one\nglobal geometry.\n\n## Layout\n\n```\nindex.csv                       one row per render, with sun and camera parameters\ndata/<site>/<tile>/images/rgb_XXX.png\ndata/<site>/<tile>/crater_field.parquet synthetic craters (x, y, diameter) in metres\ndata/<site>/<tile>/rock_field.parquet   synthetic rocks (x, y, size, rotation) in metres\ndata/<site>/<tile>/params.json          full per-case render parameters\n```\n\nThe crater and rock fields are Parquet rather than CSV: they run to millions\nof rows per tile, and Parquet with float32 columns is about 4.2x smaller than\nthe equivalent CSV (1.52 GB down to 0.36 GB across the release). float32 keeps\nroughly 0.5 mm of precision on coordinates spanning 8-1","encoding":"utf8","key":"README.md","name":"README.md","size":12304,"truncated":true}],"sourceObjectKey":"data/site04/lb/crater_field.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"hem436/lunar-south-pole-trn-dataset\")\nprint(ds)","starterSamplePath":"data/lm7/lb/images/rgb_000.png","totalObjects":1442},"summary":"1,380 path-traced orthographic renders of lunar south polar terrain for terrain relative navigation (TRN), map-relative localization and illumination-invariant feature matching.","title":"Lunar South Pole Synthetic Orthographic Renders for Terrain Relative Navigation (LOLA DEMs, 60-Sun Grid)","updatedAt":"2026-11-18T03:58:02.448Z","visibility":"public","websiteUrl":"https://github.com/nasa-jpl/lunasynth"}}