Skip to content
Other

CanViT-B COCO detector

активное зрение для детекции объектов

1 min readupdated

CanViT-B — это active-vision foundation model от проекта CanViT: Canvas Vision Transformer, который смотрит на сцену через последовательность glimpse и запоминает увиденное на общем canvas. Версия с CenterNet-головой обучена end-to-end для object detection на COCO 2017 и предсказывает боксы после каждого glimpse. Размер — около 0,1 млрд параметров, реализация на JAX / Flax NNX.

Ключевые особенности

  • Активное зрение: сцена обрабатывается серией glimpse, а состояние накапливается на canvas.
  • CenterNet-голова на canvas: декодирование боксов, оценок и индексов классов COCO.
  • JAX / Flax NNX и класс CanViTForObjectDetection.
  • 80 классов COCO; сцены 512 px, glimpse 128 px, canvas 32×32.
  • Обучение end-to-end на COCO 2017, чекпоинт после 16 172 шагов.

Пример

import flax.nnx as nnx
from canvit_nnx.detect.model import CanViTForObjectDetection
model = nnx.view(CanViTForObjectDetection.from_pretrained(
    'canvit/canvitb16-add-vpe-detect-g128px-s512px-coco-2026-08-01-nnx'
), use_running_average=True)
ПодробнееCOCO val2017, C2F: AP 0.3242, AP50 0.5334, AP75 0.3337 при 20 glimpse; AP растёт с 0.1931 на старте.
Обучение: 16 172 шагов, по 1024 сцены COCO train2017 на шаг, 10 glimpse за rollout.
Требования к VRAM в карточке не указаны; это research-чекпоинт на JAX / Flax NNX.
Лицензия MIT; претрейн — дистилляция из DINOv3.
Discuss Telegram 0 26

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.4 GB · ready files0.9 GB
  • BF16weights 0.2 GB0.7 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 · 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.

    # canvit/canvitb16-add-vpe-detect-g128px-s512px-coco-2026-08-01-nnx
    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 canvit/canvitb16-add-vpe-detect-g128px-s512px-coco-2026-08-01-nnx --local-dir ~/models/canvitb16-add-vpe-detect-g128px-s512px-coco-2026-08-01-nnx
    # 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 canvitb16-add-vpe-pretrain-g128px-s512px-in21k-dv3b16-2026-02-02 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 1All on Hugging Face
Fine-tunes 12All on Hugging Face

About the model

canvit/canvitb16-add-vpe-detect-g128px-s512px-coco-2026-08-01-nnx
Author
canvit
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

260

Ask AI about this model

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

Comments

Other

AISniff

AISniff