Project AURIONAI Use-Cases Research & Innovation
NVIDIA
visionfamily

SegFormer

NVIDIA simple and efficient transformer for semantic segmentation with a hierarchical Mix Transformer encoder and an MLP decoder.

NVIDIANVIDIAUSA
Figure from the SegFormer paper: Effective Receptive Field (ERF) on Cityscapes (average over 100 images).
Levelfamily
Classvision
Parametersnot in the notes
Contextnot in the notes
Released31 May 2021
Opennessopen weights restricted
Runs onmobile or NPU
Statusactive

Lineage

NVIDIASegFormer-B0 belongs to this familyvision31 May 20213.8M
NVIDIASegFormer-B5 belongs to this familyvision31 May 202185M
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. Confidence and evidence are the notes' own.

R&D Topics

Semantic, instance and panoptic segmentation architecturesexemplar

Identity

Developer
NVIDIA
Family
not in the notes
Released
31 May 2021 sourcesannouncement page
Weights published
2 Mar 2022
Record checked
12 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
dense transformer
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Note
MiT-B0 to B5 on ADE20K and Cityscapes; NVIDIA TAO CitySemSegFormer and PeopleSemSegFormer.
Context length
not in the notes
Input
image
Output
image
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights restricted
Licence
not in the notes sources
Access routes
open weights download, cloud marketplace
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
mobile or NPU
Footprint
Edge or toolkit builds listed in: Qualcomm AI Hub, TI edge AI model zoo.
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Benchmark Results

The notes record no benchmark result for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. arxiv.org/abs/2105.15203 https://arxiv.org/abs/2105.15203
  2. huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512 https://huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512
  3. ai.azure.com/catalog/models?task=image-segmentation https://ai.azure.com/catalog/models?task=image-segmentation
  4. huggingface.co/nvidia/segformer-b5-finetuned-ade-640-640 https://huggingface.co/nvidia/segformer-b5-finetuned-ade-640-640
  5. aihub.qualcomm.com/models/segformer_base https://aihub.qualcomm.com/models/segformer_base
  6. github.com/huggingface/transformers/tree/main/src/transf... https://github.com/huggingface/transformers/tree/main/src/transformers/models/segformer
  7. catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/citysemsegformer https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/citysemsegformer

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "SegFormer", https://aurion.semai.id/models/segformer/, records of October 2026.