Skip to content
Other

GeoSeg-PV models

weights for solar panel detection in satellite imagery

2 min readupdated

A collection of trained weights for segmenting (splitting an image into objects) individual photovoltaic modules in ultra-high-resolution satellite imagery. The models are based on various computer vision architectures (such as UNetFormer, FT-UNetFormer, DCSwin, and MANet) with fine-tuning module (FOM) support. The tool outputs frame-by-frame masks defining the locations of solar panels relative to the background. This development is useful for researchers, engineers, and geospatial data specialists analyzing solar energy and assessing rooftop potential.

Key features

  • 10 different network architectures with and without the FOM module.
  • Training on high-quality Google satellite imagery with a resolution of about 0.1 m.
  • High F1 (up to 0.952) and IoU (up to 0.908) metric scores for the best configurations.
  • Open-source Creative Commons Attribution 4.0 (CC BY 4.0) license for commercial and non-commercial use.

What you can do

  • Run inference on satellite images for automated solar panel detection.
  • Average predictions from multiple models or splits to improve segmentation accuracy.
  • Integrate weights into geospatial pipelines based on the GeoSeg-PV repository.

Example

from huggingface_hub import hf_hub_download
from geoseg.pv import load_model, predict_proba

path = hf_hub_download('nikolukac/GeoSeg-PV-models', 'ftunetformer_fom/fold0.safetensors')
model = load_model(path, device='cuda')
pv = predict_proba(model, rgb, device='cuda') >= 0.5   # rgb: uint8 array (H, W, 3), any size
Read moreDetails
• Metrics: the best ftunetformer_fom configuration shows an F1 of 0.952 ± 0.008 and an IoU of 0.908 ± 0.014.
• Requirements: GPU operation via CUDA is supported, input data consists of 8-bit RGB images normalized to ImageNet standards.
• Training details: AdamW optimizer with Lookahead, soft cross-entropy and Dice loss functions.
• Size: 85 files, total size 11.2 GB.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.4 GB · ready files0.9 GB
  • FP8 / INT8recommendedweights 0.2 GB0.7 GB
  • 4-bitweights 0.1 GB0.6 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # nikolukac/GeoSeg-PV-models
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download nikolukac/GeoSeg-PV-models --local-dir ~/models/GeoSeg-PV-models
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to GeoSeg-PV-models
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of GeoSeg-PV-models (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

nikolukac/GeoSeg-PV-models
Author
nikolukac
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments