Skip to content
Other

HRNet-W32 Footprint

маска зданий из аэрофото

2 min read

Чекпоинт для бинарной семантической сегментации контуров зданий на высокодетальных ортофото. Архитектура: HRNet-W32 + FCNHead, обучение и инференс в MMSegmentation 1.x. Это модель, которую использует пайплайн Footprint-Creater для превращения маски в GIS-ready Shapefiles.

Ключевые особенности

  • HRNet-W32 сохраняет высокое разрешение по всей сети — лучше держит границы зданий, чем агрессивные encoder-decoder схемы.
  • 2 класса: background и building; вход 1024×1024 px RGB, ImageNet mean/std, BGR→RGB.
  • Вес чекпоинта ~231 MB; inference в режиме whole.
  • Для полных ортофото — скользящее окно с ~50% перекрытием и склейка, а не подача всего снимка.
  • Семантика, не инстансы: соприкасающиеся здания сливаются в один полигон.
  • Сырая маска — не картографический продукт: в референсном пайплайне фильтрация компонент <100 м², заливка дыр, сглаживание, Douglas-Peucker и замена на min bounding rectangle при IoU выше порога.

Пример

from mmseg.apis import init_model, inference_model

# cfg — конфиг EncoderDecoder/HRNet/FCNHead из карточки
model = init_model(cfg, "iter_40000.pth", device="cuda:0")
result = inference_model(model, "patch.png")
mask = result.pred_sem_seg.data.cpu().numpy()[0]  # 0 = фон, 1 = здание
ПодробнееОбучение: CrossEntropy + Dice, SGD с PolyLR, 40 000 итераций, патчи 1024×1024 со stride 512.
Публичных бенчмарков нет: датасет приватный, IoU на публичных наборах не заявлен.
Ограничения: чёрные/белые края ортофото систематически принимаются за здания — маскируйте их до инференса; сезонность, сенсоры и необычные кровли снижают точность.
Лицензия MIT, поставляется as-is.
Discuss Telegram 1

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.2 GB · ready files0.7 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 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.

    # Uve1048/hrnet-w32-footprint
    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 Uve1048/hrnet-w32-footprint --local-dir ~/models/hrnet-w32-footprint
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to hrnet-w32-footprint
  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 hrnet-w32-footprint (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

Uve1048/hrnet-w32-footprint
Author
Uve1048
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

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

Comments

Other

OpenHeart

OpenHeart