---
license: cc-by-sa-4.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: legacysurvey_object_id
    dtype: string
  - name: astropt_15m_hsc
    list: float32
  - name: astropt_15m_legacysurvey
    list: float32
  - name: astropt_95m_hsc
    list: float32
  - name: astropt_95m_legacysurvey
    list: float32
  - name: astropt_850m_hsc
    list: float32
  - name: astropt_850m_legacysurvey
    list: float32
  - name: convnext_nano_hsc
    list: float32
  - name: convnext_nano_legacysurvey
    list: float32
  - name: convnext_tiny_hsc
    list: float32
  - name: convnext_tiny_legacysurvey
    list: float32
  - name: convnext_base_hsc
    list: float32
  - name: convnext_base_legacysurvey
    list: float32
  - name: convnext_large_hsc
    list: float32
  - name: convnext_large_legacysurvey
    list: float32
  - name: dino_small_hsc
    list: float32
  - name: dino_small_legacysurvey
    list: float32
  - name: dino_base_hsc
    list: float32
  - name: dino_base_legacysurvey
    list: float32
  - name: dino_large_hsc
    list: float32
  - name: dino_large_legacysurvey
    list: float32
  - name: dino_giant_hsc
    list: float32
  - name: dino_giant_legacysurvey
    list: float32
  - name: ijepa_huge_hsc
    list: float32
  - name: ijepa_huge_legacysurvey
    list: float32
  - name: ijepa_giant_hsc
    list: float32
  - name: ijepa_giant_legacysurvey
    list: float32
  - name: vit_base_hsc
    list: float32
  - name: vit_base_legacysurvey
    list: float32
  - name: vit_large_hsc
    list: float32
  - name: vit_large_legacysurvey
    list: float32
  - name: vit_huge_hsc
    list: float32
  - name: vit_huge_legacysurvey
    list: float32
  splits:
  - name: train
    num_bytes: 13556703431
    num_examples: 101725
  download_size: 15771223711
  dataset_size: 13556703431
---
# Legacy Survey ↔ HSC Embeddings (The Platonic Universe)

Precomputed **cross-survey embeddings** for matched sources in **Legacy Survey** and **HSC**.  
Each row is one object with multiple backbone embeddings for **both** surveys (paired by suffixes `_legacysurvey` and `_hsc`).

**Examples of columns (see Viewer for full list):**
- AstroPT: `astropt_15m_hsc`, `astropt_15m_legacysurvey`, `astropt_95m_*`, `astropt_850m_*`
- ConvNeXt: `convnext_nano_*`, `convnext_tiny_*`, `convnext_base_*`, `convnext_large_*`
- DINOv2: `dino_small_*`, `dino_base_*`, `dino_large_*`, `dino_giant_*`
- I-JEPA: `ijepa_huge_*`, `ijepa_giant_*`
- ViT: `vit_base_*`, `vit_large_*`, `vit_huge_*`
- Identifier: `legacysurvey_object_id`  


---

## Load in Python

```python
from datasets import load_dataset
import numpy as np

ds = load_dataset("UniverseTBD/legacysurvey_hsc_embeddings", split="train")
print("Columns:", ds.column_names[:10], "...")
row = ds[0]
```