Segment Anything (SAM + MobileSAM) — ONNX Models

ONNX-exported versions of Meta's Segment Anything Model (SAM) and MobileSAM, ready for CPU/GPU inference with ONNX Runtime — no PyTorch required at runtime.

These models are used by AnyLabeling for AI-assisted image annotation, and exported by samexporter.

Available Models

File	Variant	Encoder size	Notes
`sam_vit_b_01ec64.zip`	SAM ViT-B	~90 MB	Fastest, lowest accuracy
`sam_vit_b_01ec64_quant.zip`	SAM ViT-B (Quant)	~25 MB	Quantized — smaller & faster
`sam_vit_l_0b3195.zip`	SAM ViT-L	~330 MB	Good balance
`sam_vit_l_0b3195_quant.zip`	SAM ViT-L (Quant)	~83 MB	Quantized — smaller & faster
`sam_vit_h_4b8939.zip`	SAM ViT-H	~630 MB	Highest accuracy
`sam_vit_h_4b8939_quant.zip`	SAM ViT-H (Quant)	~158 MB	Quantized — smaller & faster
`mobile_sam_20230629.zip`	MobileSAM	~9 MB	Ultra-lightweight

Each zip contains two ONNX files: an encoder (runs once per image) and a decoder (runs interactively for each prompt).

Prompt Types

Point (+point / -point): click to include/exclude regions
Rectangle: draw a bounding box around the target object

Use with AnyLabeling (Recommended)

AnyLabeling is a desktop annotation tool with a built-in model manager that downloads, caches, and runs these models automatically — no coding required.

Install: pip install anylabeling
Launch: anylabeling
Click the Brain button → select a SAM model from the dropdown
Use point or rectangle prompts to segment objects

Use Programmatically with ONNX Runtime

import urllib.request, zipfile, pathlib
# Download and extract
url = "https://huggingface.co/vietanhdev/segment-anything-onnx-models/resolve/main/sam_vit_b_01ec64.zip"
urllib.request.urlretrieve(url, "sam_vit_b_01ec64.zip")
with zipfile.ZipFile("sam_vit_b_01ec64.zip") as z:
    z.extractall("sam_vit_b_01ec64")

Then use samexporter's inference module:

pip install samexporter
python -m samexporter.inference \
    --encoder_model sam_vit_b_01ec64/sam_vit_b_encoder.onnx \
    --decoder_model sam_vit_b_01ec64/sam_vit_b_decoder.onnx \
    --image photo.jpg \
    --prompt prompt.json \
    --output result.png

Re-export from Source

To re-export or customize the models using samexporter:

pip install samexporter
# Export SAM ViT-H encoder + decoder
python -m samexporter.export_encoder \
    --checkpoint original_models/sam_vit_h_4b8939.pth \
    --output output_models/sam_vit_h_4b8939.encoder.onnx \
    --model-type vit_h --use-preprocess
python -m samexporter.export_decoder \
    --checkpoint original_models/sam_vit_h_4b8939.pth \
    --output output_models/sam_vit_h_4b8939.decoder.onnx \
    --model-type vit_h --return-single-mask
# Or convert all SAM variants at once:
bash convert_all_meta_sam.sh

Related Repositories

Repo	Description
vietanhdev/samexporter	Export scripts, inference code, conversion tools
vietanhdev/anylabeling	Desktop annotation app powered by these models
facebookresearch/segment-anything	Original SAM by Meta
ChaoningZhang/MobileSAM	Original MobileSAM

License

The ONNX models are derived from Meta's SAM and MobileSAM, both released under the Apache 2.0 license. The export code is part of samexporter, released under the MIT license.

Downloads last month: -; Downloads are not tracked for this model. How to track