Skip to content
Other

Argus V3

improved ink detection model for papyrus scans

2 min readupdated

Argus V3 is built to detect ink on native (raw) 9-micrometer scroll scans and serves as a direct replacement for the Reader V3 architecture. It uses the same network and file structure, making it easy to drop into HerculeX, ink-sdk, and other tools. The model delivers higher accuracy on native benchmarks than its predecessors and is well-suited for researchers working on ancient texts in the Vesuvius Challenge.

Key features

  • Outperforms on benchmarks: beats Reader V3 on AUC, bAcc, and Dice metrics across all five native test segments.
  • Weight averaging: created as a uniform average of four fine-tunes of the base domenicor046/reader-v3 model.
  • Full compatibility: uses the exact same schema and file layout as reader-v3.pth, acting as a drop-in replacement.
  • Fewer false positives: less likely to misclassify background noise as ink across most test data.

What you can do

  • Swap into HerculeX: use the model file as a quick drop-in replacement for Reader V3 via an environment variable.
  • Analyze native scans: detect hidden ink on native 9 µm papyrus scans.
  • Tune depth windows: average windows shifted by -2, 0, and +2 layers for an extra boost in accuracy.

Example

Example of setting the model weights path in HerculeX:

HERCULEX_WEIGHTS=/path/to/argus-v3.pth
Read moreDetails
Metrics (native tests): AUC up to 0.924, bAcc@0.5 up to 0.835, Dice up to 0.728.
Limitations: on official validation masks (merged 2.4 µm renders), Reader V3 and Argus V1 score slightly higher on AUC and Dice.
Training: 12,000 iterations, batch size 8, SGD, BCE loss with 0.5 label smoothing plus Dice, dataset of random 30% blocks from 70 native segments.
License: MIT.
Examples
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.3 GB · ready files0.8 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.

    # ruiconte/argus-v3
    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 ruiconte/argus-v3 --local-dir ~/models/argus-v3
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to argus-v3
  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 argus-v3 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

ruiconte/argus-v3
Author
ruiconte
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

Language models

Argus v5

Argus v5
Other

Connais 1

Connais 1