Skip to content
Other

f1_fight_mnv3_16f

fight detector for video surveillance systems

2 min readupdated

The f1_fight_mnv3_16f model solves the task of automatically recognizing fights and conflicts in video clips. It is built on the lightweight MobileNetV3-Large architecture with added temporal convolution layers and analyzes 16-frame sequences. The development is designed for integration into smart security systems and covers predictive monitoring tasks in public spaces. The model stands out for its compact size (3.91 million parameters) and high inference speed on standard hardware.

Key features

  • Trained on the RWF-2000 dataset of real street conflicts.
  • Very fast: 16.2 ms on CPU and 3.9 ms on an RTX 5070.
  • Available in ONNX format (16 MB) and PyTorch.
  • Accepts NCTHW-format clips at 224x224 resolution as input.

What you can do

  • Integrate the model into a smart video surveillance system for real-time automatic detection of scuffles.
  • Use it as an analytics component in security robots or drones, taking camera angles into account.
  • Set up a two-stage filtering pipeline to reduce false positives.

Example

# Загрузка модели и инференс через скрипт репозитория
python scripts/fetch_model.py --model f1_fight_mnv3_16f
Read moreDetails
Metrics: validation F1 is 0.824, AUC is 0.859.
Hardware: 3.91 M parameters, ~16.0 MB in ONNX.
Limitations: high sensitivity to camera angle and domain (recall drops on sports videos or non-standard angles), backend confirmation required to avoid false alarms.
Base model: MobileNetV3-Large (ImageNet).
Discuss Telegram 0

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.

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

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

About the model

kuccy/patrol-detector
Author
kuccy
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