Skip to content
Other

MobileSAM-ONNX: Accelerated ONNX graphs for image segmentation

2 min readupdated

A set of weights and graphs in the ONNX format for MobileSAM v2 (a lightweight version of the Segment Anything image segmentation model), prepared by the moonbox-dev team for running on devices using ONNX Runtime (CPU, CUDA, DirectML). The model consists of a compact TinyViT-5M image encoder and a mask decoder, replacing the heavy components of the original for high performance on mobile and local devices. The solution comes with a ready-to-use pipeline.py file for image preprocessing and prompt coordinate processing. This new release is useful for developers of mobile apps, local services, and automated data labeling systems.

Key features

  • Lightweight TinyViT-5M encoder (just 9.8M parameters in total)
  • ONNX Runtime support for CPU, CUDA, and DirectML
  • fp16 precision with minimal mask deviation by the IoU metric
  • Built-in pipeline.py script for scaling, normalization, and working with points/boxes

What you can do

  • Integrate the segmentation model into mobile or local desktop apps via ONNX Runtime
  • Quickly generate object masks from points or bounding boxes on user photos

Example

import sys, numpy as np
from PIL import Image
sys.path.insert(0, "fp16")
import pipeline
sam = pipeline.MobileSAM("fp16")
emb, size = sam.embed(np.array(Image.open("photo.jpg").convert("RGB")))
masks, scores = sam.point(emb, size, [[410, 300]])          # 3 masks [3, H, W] (bool)
box_mask, score = sam.box(emb, size, [120, 80, 700, 560])
Read moreDetails
Benchmark: average mask IoU 0.998 / worst 0.977 for the fp16 build
File size: encoder 14.1 MB, decoder 16.5 MB
Dependencies: numpy, onnxruntime (or onnxruntime-gpu), Pillow
License: apache-2.0
Examples
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.

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

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

About the model

moonbox-dev/MobileSAM-ONNX
Author
moonbox-dev
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