AIBOX Application Case - Background Removal via U²-Net

AIBOX Series

Both AIBOX-OrinNano and AIBOX-OrinNX are equipped with NVIDIA's original Jetson Orin core board module, standard industrial-grade full-metal shell, aluminum alloy structure for heat conduction, and a banner grille design on the side of the top cover shell for efficient heat dissipation, ensuring computing performance and stability under high-temperature operation, and meeting various industrial-grade application requirements.

Application Case: Background Removal

Background Removal technology has become an important tool in the field of image processing, and is mainly used in scenarios such as image editing, data analysis, and application development.

Typical applications:

• Image processing: e-commerce product image extraction, portrait beautification, medical image analysis.
• Video processing: real-time green screen replacement, dynamic object tracking.
• Scientific research preprocessing: improving quantification accuracy by background removal in meteorological chromatographic analysis.

U²-Net

U²-Net (U-squared Net) is a deep learning-based image segmentation model designed for high-precision background removal tasks. Its core technical features and application scenarios are as follows:

• Dual U-shaped codec structure.
• Deep supervision and loss function.
• Lightweight design.

Download Source Code

1. $ git clone --recursive --depth=1
2. https://github.com/dusty-nv/jetson-inference

Compile/Install

Reference: https://github.com/dusty-nv/jetson-inference/blob/master/docs/building-repo-2.md

Running the Example

1. # remove the background (with alpha)
2. $ ./backgroundnet.py images/bird_0.jpg images/test/bird_mask.png
4. # replace the background
5. $ ./backgroundnet.py --replace=images/snow.jpg images/bird_0.jpg images/test/bird_replace.jpg