Skip to content
Other

Lunar crater detector

lunar crater detector on lunar images

2 min readupdated

The model is a specialized object detector based on the YOLO11s architecture, trained to find lunar craters ranging from 2 to 30 kilometers in diameter. It is built on real tiles from the global LRO WAC (Lunar Reconnaissance Orbiter Wide-Angle Camera) mosaic with a resolution of 100 meters per pixel and the Robbins 2018 catalog. The tool automatically determines coordinates and confidence for each detected crater in orbital images of the Moon.

Key features

  • Training on real data: LRO WAC tiles and annotations from the authoritative Robbins 2018 lunar crater database were used.
  • Good metrics: On a test set from previously unilluminated longitude ranges, the model achieved an mAP50 of 0.788.
  • Regional limitation: Trained for latitudes up to 45 degrees and orthographic orbital images.

What you can do

  • Orbital image analysis: Find and label craters on new lunar surface tiles at a 640-pixel resolution.
  • Automate planetary science tasks: Use the model to process large arrays of space images and catalog impact structures.

Example

from huggingface_hub import hf_hub_download
from ultralytics import YOLO

model = YOLO(hf_hub_download("mvdelibas/ay-krater-lro-wac-yolo11s-v1", "ay_krater_yolo11s.pt"))
result = model.predict("moon_tile.jpg", imgsz=640, conf=0.25)[0]
print(result.boxes.xyxy, result.boxes.conf)
Read moreDetails
Metrics (test): mAP50 = 0.788, mAP50-95 = 0.531, Precision = 0.731, Recall = 0.697.
Limitations: craters smaller than 2 km are not labeled in the catalog (may be falsely detected as False Positives), the model does not work with polar regions, Martian images, and rover perspectives.
Dataset: about 3,000 tiles (121 thousand boxes), 60 training epochs.
Discuss Telegram 0 10

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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.

    # mvdelibas/ay-krater-lro-wac-yolo11s-v1
    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 mvdelibas/ay-krater-lro-wac-yolo11s-v1 --local-dir ~/models/ay-krater-lro-wac-yolo11s-v1
    # The model is in ~/models. How to run it is in the model card (link above).
  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 ay-krater-lro-wac-yolo11s-v1 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

mvdelibas/ay-krater-lro-wac-yolo11s-v1
Author
mvdelibas
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

100

Ask AI about this model

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

Comments