Skip to content
Other

Anticipa colisiones

прогноз ДТП по 2 секундам dashcam

2 min read

Академический проект Pontificia Universidad Javeriana на базе MCG-NJU/videomae-base-finetuned-kinetics. По окну 2 с (16 равномерных кадров) модель выдаёт оценку 0–1 того, что через 0,5–1,5 с после конца окна произойдёт столкновение или почти столкновение. Это не официальная модель Nexar, использование некоммерческое.

Ключевые особенности

  • Вход — 2 с видео: 16 кадров в хронологическом порядке; модель никогда не видит сам момент удара.
  • База VideoMAE; дообучены последние 4 блока encoder, fc_norm и новая бинарная голова.
  • 28 350 721 обучаемых параметров из 86 227 969; обучение на 2 836 окнах, seed 42.
  • Порог приложения 0,0781 — максимум F1 на валидации; он специфичен только для этой модели.
  • Вход pixel_values [n, 16, 3, 224, 224], геометрия squash без обрезки кадра.

Пример

import torch
from transformers import AutoVideoProcessor, VideoMAEForVideoClassification
model = VideoMAEForVideoClassification.from_pretrained("Jeol28/anticipa-colisiones").eval()
proc = AutoVideoProcessor.from_pretrained("Jeol28/anticipa-colisiones")
x = proc(videos=[clip], return_tensors="pt", do_center_crop=False, size={"height": 224, "width": 224})["pixel_values"]
score = torch.sigmoid(model(pixel_values=x).logits[:, 0])
ПодробнееНа тесте Nexar: mAP 0,851, ROC-AUC 0,863; AP при Δ = 0,5/1,0/1,5 с — 0,890/0,859/0,804.
С порогом 0,0781 матрица: VN 452, FP 220, FN 92, VP 580; оценка не калибрована.
Есть сдвиг к городским сценам; AP падает с ростом горизонта прогноза.
На CPU Colab видео 40 с обрабатывалось 338,4 с; домен — горизонтальные 1280×720 из датасета Nexar.
Лицензия cc-by-nc-4.0; не для принятия решений о безопасности и не для автомобиля.
Discuss Telegram 0 2

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.3 GB · ready files0.9 GB
  • BF16weights 0.2 GB0.7 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 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 · F32 — 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.

    # Jeol28/anticipa-colisiones
    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 Jeol28/anticipa-colisiones --local-dir ~/models/anticipa-colisiones
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to anticipa-colisiones
  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 videomae-base-finetuned-kinetics (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 12All on Hugging Face
LoRAs and adapters 1All on Hugging Face

About the model

Jeol28/anticipa-colisiones
Author
Jeol28
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

20

Ask AI about this model

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

Comments