Rembg

Rembg is a tool to remove images background.

If this project has helped you, please consider making a donation.

Sponsors

	RemBG API https://www.rembg.com Premium background removal API – from $1/month
	withoutBG API https://withoutbg.com High-quality background removal API at affordable rates
	PhotoRoom Remove Background API https://photoroom.com/api Fast and accurate background remover API

Requirements

text python: >=3.10, <3.14

Installation

If you have onnxruntime already installed, just install rembg:

bash pip install rembg # for library pip install "rembg[cli]" # for library + cli

Otherwise, install rembg with explicit CPU/GPU support.

CPU support:

bash pip install rembg[cpu] # for library pip install "rembg[cpu,cli]" # for library + cli

GPU support (NVidia/Cuda):

First of all, you need to check if your system supports the onnxruntime-gpu.

Go to onnxruntime.ai and check the installation matrix.

If yes, just run:

bash pip install "rembg[gpu]" # for library pip install "rembg[gpu,cli]" # for library + cli

Nvidia GPU may require onnxruntime-gpu, cuda, and cudnn-devel. #668 . If rembg[gpu] doesn't work and you can't install cuda or cudnn-devel, use rembg[cpu] and onnxruntime instead.

GPU support (AMD/ROCM):

ROCM support requires the onnxruntime-rocm package. Install it following AMD's documentation.

If onnxruntime-rocm is installed and working, install the rembg[rocm] version of rembg:

bash pip install "rembg[rocm]" # for library pip install "rembg[rocm,cli]" # for library + cli

Usage as a cli

After the installation step you can use rembg just typing rembg in your terminal window.

The rembg command has 4 subcommands, one for each input type:

• i for files
• p for folders
• s for http server
• b for RGB24 pixel binary stream

You can get help about the main command using:

shell rembg --help

As well, about all the subcommands using:

shell rembg <COMMAND> --help

rembg i

Used when input and output are files.

Remove the background from a remote image

shell curl -s http://input.png | rembg i > output.png

Remove the background from a local file

shell rembg i path/to/input.png path/to/output.png

Remove the background specifying a model

shell rembg i -m u2netp path/to/input.png path/to/output.png

Remove the background returning only the mask

shell rembg i -om path/to/input.png path/to/output.png

Remove the background applying an alpha matting

shell rembg i -a path/to/input.png path/to/output.png

Passing extras parameters

```shell SAM example

rembg i -m sam -x '{ "sam_prompt": [{"type": "point", "data": [724, 740], "label": 1}] }' examples/plants-1.jpg examples/plants-1.out.png ```

```shell Custom model example

rembg i -m u2netcustom -x '{"modelpath": "~/.u2net/u2net.onnx"}' path/to/input.png path/to/output.png ```

rembg p

Used when input and output are folders.

Remove the background from all images in a folder

shell rembg p path/to/input path/to/output

Same as before, but watching for new/changed files to process

shell rembg p -w path/to/input path/to/output

rembg s

Used to start http server.

shell rembg s --host 0.0.0.0 --port 7000 --log_level info

To see the complete endpoints documentation, go to: http://localhost:7000/api.

Remove the background from an image url

shell curl -s "http://localhost:7000/api/remove?url=http://input.png" -o output.png

Remove the background from an uploaded image

shell curl -s -F file=@/path/to/input.jpg "http://localhost:7000/api/remove" -o output.png

rembg b

Process a sequence of RGB24 images from stdin. This is intended to be used with another program, such as FFMPEG, that outputs RGB24 pixel data to stdout, which is piped into the stdin of this program, although nothing prevents you from manually typing in images at stdin.

shell rembg b image_width image_height -o output_specifier

Arguments:

• image_width : width of input image(s)
• image_height : height of input image(s)
• output_specifier: printf-style specifier for output filenames, for example if output-%03u.png, then output files will be named output-000.png, output-001.png, output-002.png, etc. Output files will be saved in PNG format regardless of the extension specified. You can omit it to write results to stdout.

Example usage with FFMPEG:

shell ffmpeg -i input.mp4 -ss 10 -an -f rawvideo -pix_fmt rgb24 pipe:1 | rembg b 1280 720 -o folder/output-%03u.png

The width and height values must match the dimension of output images from FFMPEG. Note for FFMPEG, the "-an -f rawvideo -pix_fmt rgb24 pipe:1" part is required for the whole thing to work.

Usage as a library

Input and output as bytes

```python from rembg import remove

inputpath = 'input.png' outputpath = 'output.png'

with open(inputpath, 'rb') as i: with open(outputpath, 'wb') as o: input = i.read() output = remove(input) o.write(output) ```

Input and output as a PIL image

```python from rembg import remove from PIL import Image

inputpath = 'input.png' outputpath = 'output.png'

input = Image.open(inputpath) output = remove(input) output.save(outputpath) ```

Input and output as a numpy array

```python from rembg import remove import cv2

inputpath = 'input.png' outputpath = 'output.png'

input = cv2.imread(inputpath) output = remove(input) cv2.imwrite(outputpath, output) ```

Force output as bytes

```python from rembg import remove

inputpath = 'input.png' outputpath = 'output.png'

with open(inputpath, 'rb') as i: with open(outputpath, 'wb') as o: input = i.read() output = remove(input, forcereturnbytes=True) o.write(output) ```

How to iterate over files in a performatic way

```python from pathlib import Path from rembg import remove, new_session

session = new_session()

for file in Path('path/to/folder').glob('*.png'): inputpath = str(file) outputpath = str(file.parent / (file.stem + ".out.png"))

with open(input_path, 'rb') as i:
    with open(output_path, 'wb') as o:
        input = i.read()
        output = remove(input, session=session)
        o.write(output)

```

To see a full list of examples on how to use rembg, go to the examples page.

Usage as a docker

Only CPU

Just replace the rembg command for docker run danielgatis/rembg.

Try this:

shell docker run -v path/to/input:/rembg danielgatis/rembg i input.png path/to/output/output.png

Nvidia CUDA Hardware Acceleration

Requirement: using CUDA in docker needs your host has NVIDIA Container Toolkit installed. NVIDIA Container Toolkit Install Guide

Nvidia CUDA Hardware Acceleration needs cudnn-devel so you need to build the docker image by yourself. #668

Here is a example shows you how to build an image and name it rembg-nvidia-cuda-cudnn-gpu shell docker build -t rembg-nvidia-cuda-cudnn-gpu -f Dockerfile_nvidia_cuda_cudnn_gpu . Be aware: It would take 11GB of your disk space. (The cpu version only takes about 1.6GB). Models didn't included.

After you build the image, run it like this as a cli shell sudo docker run --rm -it --gpus all -v /dev/dri:/dev/dri -v $PWD:/rembg rembg-nvidia-cuda-cudnn-gpu i -m birefnet-general input.png output.png

• Trick 1: Actually you can also make up a nvidia-cuda-cudnn-gpu image and install rembg[gpu, cli] in it.
• Trick 2: Try param -v /somewhereYouStoresModelFiles/:/root/.u2net so to download/store model files out of docker images. You can even comment the line RUN rembg d u2net so when building the image, it download will no models, so you can download the specific model you want even without the default u2net model.

Models

All models are downloaded and saved in the user home folder in the .u2net directory.

The available models are:

• u2net (download, source): A pre-trained model for general use cases.
• u2netp (download, source): A lightweight version of u2net model.
• u2nethumanseg (download, source): A pre-trained model for human segmentation.
• u2netclothseg (download, source): A pre-trained model for Cloths Parsing from human portrait. Here clothes are parsed into 3 category: Upper body, Lower body and Full body.
• silueta (download, source): Same as u2net but the size is reduced to 43Mb.
• isnet-general-use (download, source): A new pre-trained model for general use cases.
• isnet-anime (download, source): A high-accuracy segmentation for anime character.
• sam (download encoder, download decoder, source): A pre-trained model for any use cases.
• birefnet-general (download, source): A pre-trained model for general use cases.
• birefnet-general-lite (download, source): A light pre-trained model for general use cases.
• birefnet-portrait (download, source): A pre-trained model for human portraits.
• birefnet-dis (download, source): A pre-trained model for dichotomous image segmentation (DIS).
• birefnet-hrsod (download, source): A pre-trained model for high-resolution salient object detection (HRSOD).
• birefnet-cod (download, source): A pre-trained model for concealed object detection (COD).
• birefnet-massive (download, source): A pre-trained model with massive dataset.

How to train your own model

If You need more fine tuned models try this: https://github.com/danielgatis/rembg/issues/193#issuecomment-1055534289

Some video tutorials

• https://www.youtube.com/watch?v=3xqwpXjxyMQ
• https://www.youtube.com/watch?v=dFKRGXdkGJU
• https://www.youtube.com/watch?v=Ai-BS_T7yjE
• https://www.youtube.com/watch?v=D7W-C0urVcQ

References

• https://arxiv.org/pdf/2005.09007.pdf
• https://github.com/NathanUA/U-2-Net
• https://github.com/pymatting/pymatting

FAQ

When will this library provide support for Python version 3.xx?

This library directly depends on the onnxruntime library. Therefore, we can only update the Python version when onnxruntime provides support for that specific version.

Buy me a coffee

Liked some of my work? Buy me a coffee (or more likely a beer)

Star History

License

Copyright (c) 2020-present Daniel Gatis

Licensed under MIT License