Batch Background Removal with PaddlePaddle: One-Click Python Script

Preparation

We first install the PaddlePaddle deep-learning framework (CPU version) using the official mirror:

python -m pip install paddlepaddle -i https://mirror.baidu.com/pypi/simple

After installation we verify it in a Python session:

import paddle.fluid
paddle.fluid.install_check.run_check()

The output ends with “Your Paddle works well … Your Paddle is installed successfully!” indicating a successful setup.

Next we install PaddleHub, which provides ready-made models:

pip install -i https://mirror.baidu.com/pypi/simple paddlehub

Code Implementation

The workflow consists of four steps: import the module, load the segmentation model, collect image files, and run the segmentation.

import os, paddlehub as hub
huseg = hub.Module(name='deeplabv3p_xception65_humanseg')  # load model
path = './imgs/'                                          # image folder
files = [path + i for i in os.listdir(path)]               # list files
results = huseg.segmentation(data={'image': files})       # perform cutout

Place the source pictures in the imgs directory. After running the script, PaddleHub saves the results in a humanseg_output folder, preserving the original filenames but with a .png extension and a white background where the subject was removed.

Result

Below are examples of the original photos and the corresponding cut-out images produced by the script.

The tool correctly isolates people (single or multiple) and outputs them with a white background; minor artifacts may appear on fine details, but the overall quality is satisfactory.

Conclusion

Using PaddlePaddle and PaddleHub, a five-line Python program can batch-process images to remove backgrounds, saving time and effort for developers who need quick visual effects.