Name: segformer b0 finetuned ade 512 512
Author: Xenova

Overview

SegFormer-B0 (finetuned on ADE20K) is a lightweight, hierarchical Transformer-based model designed for efficient semantic segmentation. Unlike traditional CNN-based encoders, it utilizes a Mix Transformer (MiT) architecture that captures multi-scale global context without requiring computationally expensive positional encodings. This specific version is optimized for 512x512 resolution, making it ideal for real-time edge deployment or integration into applications where low latency and minimal memory overhead are critical. Developers can leverage this model for scene parsing, autonomous navigation, or image masking tasks, benefiting from a strong balance between inference speed and mIoU performance compared to heavier models like DeepLabV3.

Highlights

Efficient Mix Transformer encoder for real-time semantic segmentation
Optimized for 512x512 resolution with low memory overhead
High-performance scene parsing trained on the ADE20K dataset
Apache-2.0 license ensuring flexible commercial integration
Superior inference speed compared to traditional CNN architectures

Usage

Install

# Install Hugging Face transformers
pip install transformers torch

SDK Usage

# Load model with transformers
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("Xenova/segformer-b0-finetuned-ade-512-512")
tokenizer = AutoTokenizer.from_pretrained("Xenova/segformer-b0-finetuned-ade-512-512")

Hugging Face Download

We recommend downloading the model via the Hugging Face CLI or Hub SDK.

Guidance：Before downloading, install huggingface_hub with:

Guidance

pip install -U huggingface_hub

CLI Download

Download the full repository

huggingface-cli download Xenova/segformer-b0-finetuned-ade-512-512

Download a single file to a local folder (e.g. config.json into ./dir)

huggingface-cli download Xenova/segformer-b0-finetuned-ade-512-512 config.json --local-dir ./dir

See the official docs for more CLI options

SDK Download

# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('Xenova/segformer-b0-finetuned-ade-512-512')

Git Download

Make sure git-lfs is installed first

Git Download

git lfs install
git clone https://huggingface.co/Xenova/segformer-b0-finetuned-ade-512-512

To skip LFS large-file downloads, use:

Skip LFS

GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/Xenova/segformer-b0-finetuned-ade-512-512

Model files are hosted on the Hugging Face Hub — download directly via HF CLI / SDK / Git, not through this site.

PyTorch / Transformers Usage

Install Transformers

pip install -U transformers torch

Load the model and run inference

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('Xenova/segformer-b0-finetuned-ade-512-512')
tokenizer = AutoTokenizer.from_pretrained('Xenova/segformer-b0-finetuned-ade-512-512')

Model Download

We recommend downloading the model via the ModelScope CLI or SDK.

Guidance：Before downloading, install ModelScope with:

Guidance

pip install modelscope

CLI Download

Download the full repository

modelscope download --model Xenova/segformer-b0-finetuned-ade-512-512

Download a single file to a local folder (e.g. README.md into ./dir)

modelscope download --model Xenova/segformer-b0-finetuned-ade-512-512 README.md --local_dir ./dir

See the docs for more CLI options

SDK Download

# 模型下载
from modelscope import snapshot_download
model_dir = snapshot_download('Xenova/segformer-b0-finetuned-ade-512-512')

Git Download

Make sure git-lfs is installed first

Git Download

git lfs install
git clone https://www.modelscope.cn/Xenova/segformer-b0-finetuned-ade-512-512.git

To skip LFS large-file downloads, use:

Skip LFS

GIT_LFS_SKIP_SMUDGE=1 git clone https://www.modelscope.cn/Xenova/segformer-b0-finetuned-ade-512-512.git

ModelScope 模型页直接下载模型文件；无需将模型文件放在本站服务器。

Notebook Quickstart

Install the ModelScope library

pip install "modelscope[audio,cv,nlp,multi-modal,science]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

Load the model and run inference

from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

p = pipeline('text-generation', 'Xenova/segformer-b0-finetuned-ade-512-512')

Full Documentation

来源: HuggingFace

---
base_model: nvidia/segformer-b0-finetuned-ade-512-512
library_name: transformers.js
pipeline_tag: image-segmentation
---

https://huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512 with ONNX weights to be compatible with Transformers.js.

Usage (Transformers.js)

If you haven't already, you can install the Transformers.js JavaScript library from NPM using:

bash

npm i @huggingface/transformers

Example: Image segmentation with Xenova/segformer-b0-finetuned-ade-512-512.

js

import { pipeline } from '@huggingface/transformers';
// Create an image segmentation pipeline
const segmenter = await pipeline('image-segmentation', 'Xenova/segformer-b0-finetuned-ade-512-512');
// Segment an image
const url = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/house.jpg';
const output = await segmenter(url);
console.log(output)
// [
//   {
//     score: null,
//     label: 'wall',
//     mask: RawImage { ... }
//   },
//   {
//     score: null,
//     label: 'building',
//     mask: RawImage { ... }
//   },
//   ...
// ]

You can visualize the outputs with:

js

for (const l of output) {
  l.mask.save(${l.label}.png);
}

---

Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using 🤗 Optimum and structuring your repo like this one (with ONNX weights located in a subfolder named onnx).