SAVRN
Search Contact SAVRN

Open-weight model

vitmatte-small-composition-1k

by HUST Vision Lab hustvl/vitmatte-small-composition-1k

ViTMatte model trained on Composition-1k. It was introduced in the paper ViTMatte: Boosting Image Matting with Pretrained Plain Vision Transformers by Yao et al. and first released in this repository.

Parameters26M
Context
Weights206.6 MB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads733k

Runs On

What it takes to serve vitmatte-small-composition-1k (26M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.1 GB 0.1 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.0 GB 0.0 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.0 GB 0.0 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on vitmatte-small-composition-1k

One tenth of a gigabyte covers this model at 16-bit, weights and working memory. ViTMatte from HUST Vision Lab does image matting, estimating the foreground object in an image, with 26 million parameters: a Vision Transformer with a lightweight head on top, trained on Composition-1k. The listed cheapest setup is one MI300X at $1.85 an hour, but 0.1 GB on a 192 GB card is a rounding error; run it beside whatever already handles your image pipeline.

Apache 2.0 permits commercial use, modification and redistribution, requires the license and NOTICE file to stay attached and significant changes to be stated, and includes an express patent grant from contributors. The model card was not written by the team that released it, so the paper, arXiv 2305.15272, is the primary reference, and with no context length or task label listed, test it on your own images first.

Model Card

By HUST Vision Lab, published under apache-2.0, revision 6a58ad764640.

ViTMatte model

ViTMatte model trained on Composition-1k. It was introduced in the paper ViTMatte: Boosting Image Matting with Pretrained Plain Vision Transformers by Yao et al. and first released in this repository.

Disclaimer: The team releasing ViTMatte did not write a model card for this model so this model card has been written by the Hugging Face team.

Model description

ViTMatte is a simple approach to image matting, the task of accurately estimating the foreground object in an image. The model consists of a Vision Transformer (ViT) with a lightweight head on top.

ViTMatte high-level overview. Taken from the original paper.

Intended uses & limitations

You can use the raw model for image matting. See the model hub to look for other fine-tuned versions that may interest you.

How to use

We refer to the docs.

BibTeX entry and citation info

Read the full model card (169 words)

Configuration

Architecture
VitMatteForImageMatting
Hidden size
384
Stored precision
float32
Model type
vitmatte

Identity and Version

Repository
hustvl/vitmatte-small-composition-1k
Publisher
HUST Vision Lab
Task
Not stated by the source
Modality
Other
Library
transformers
Parameters
26M parameters
Languages
Not stated by the source
Revision
6a58ad7646403c1df626fbd746900aec7361ea1d
First published
2023-09-10
Last updated
2024-03-29

Files and Weights

6 files, 206.6 MB in total. The weights are 2 files totalling 206.6 MB in bin, safetensors.

Weights2 files · 206.6 MB
Configuration2 files · 1.1 KB
Documentation1 file · 1.7 KB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
model.safetensorsWeights103.3 MB bda9289db1bb
pytorch_model.binWeights103.3 MB 6ec6aed44bc8
config.jsonConfiguration837 B
preprocessor_config.jsonConfiguration284 B
README.mdDocumentation1.7 KB
.gitattributesRepository1.5 KB

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
206.6 MB
Download from HUST Vision Lab

Released by HUST Vision Lab through its official repository on Hugging Face. Read the license.

Built From

  • Described by arXiv:2305.15272

Memory Requirements

PrecisionWeights in memory
As published206.6 MB
16-bit0.1 GB
8-bit0.0 GB
4-bit0.0 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About vitmatte-small-composition-1k

How much GPU memory does vitmatte-small-composition-1k need?

About 0.1 GB at 16-bit and 0 GB at 4-bit: the weights (26M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run vitmatte-small-composition-1k on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use vitmatte-small-composition-1k commercially?

Yes. vitmatte-small-composition-1k is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.