Downloads · 30 days
11
13% of all-time downloads
betaearth-segformer-film-scratch
asterisk-labs/betaearth-segformer-film-scratch
betaearth-segformer-film-scratch is a feature extraction model from asterisk-labs. Use it when you need embeddings to search or compare text. It is set up for betaearth. The card lists the license as cc-by-4.0.
BetaEarth SegFormer-B2 from scratch+FiLM — fully trainable end-to-end
All-time downloads
85
Public
Repo size
420 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt420 MB · 100%
At a glance
- Task
- Feature Extraction
- Library
- betaearth
- License
- cc-by-4.0
- Model type
- betaearth-segformer
- Access
- Public
- Created
- Apr 6, 2026
- Updated
- Apr 6, 2026
- SHA
- adbbde0d
Model card
From the Hugging Face model README
betaearth-segformer-film-scratch
BetaEarth SegFormer-B2 from scratch+FiLM — fully trainable end-to-end
Part of the BetaEarth family — fully trainable, with FiLM day-of-year conditioning.
| Metric | Value |
|---|---|
| Test cosine similarity | 0.883 |
| LULC downstream accuracy | 0.835 |
| Trainable parameters | 104.8M |
| Total parameters | 104.8M |
| Inputs | S2 L1C+L2A (9ch), S1 RTC (2ch), COP-DEM (1ch), DOY |
| Output | (H, W, 64) float32, L2-normalised |
Usage
pip install betaearth
from betaearth import BetaEarth
model = BetaEarth.from_pretrained("asterisk-labs/betaearth-segformer-film-scratch")
embedding = model.predict(
s2_l2a=s2_l2a, # (9, H, W) uint16
s1=s1, # (2, H, W) float32
dem=dem, # (1, H, W) float32
doy=182,
)
# embedding: (H, W, 64) numpy array
All BetaEarth models
| Model | Cos Sim | Params | Best for |
|---|---|---|---|
| betaearth-segformer-film | 0.886 | 0.3M | Best quality |
| betaearth-segformer-film-hilr | 0.886 | 0.3M | Alt frozen |
| betaearth-segformer | 0.880 | 104.8M | No timestamp |
| betaearth-segformer-film-scratch | 0.883 | 104.8M | End-to-end |
| betaearth-rgb-only | 0.836 | 26.3M | Minimal data |
Citation
@inproceedings{czerkawski2026betaearth,
title = {BetaEarth: Emulating Closed-Source Earth Observation Foundation Models Through Their Public Embeddings},
author = {Czerkawski, Mikolaj},
booktitle = {ISPRS Congress 2026},
year = {2026}
}
License
CC-BY 4.0. Training data attribution: "The AlphaEarth Foundations Satellite Embedding dataset is produced by Google and Google DeepMind."
- Task
- Feature Extraction
- Library
- betaearth
- Type
- betaearth-segformer
- License
- cc-by-4.0
- Created
- Apr 6, 2026
- Updated
- Apr 6, 2026