u2net

U^2-Net

input

(Image from https://github.com/NathanUA/U-2-Net/blob/master/test_data/test_images/girl.png)

• Ailia input shape: (1, 3, 320, 320)

output

usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 u2net.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 u2net.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 u2net.py --video VIDEO_PATH

You can select a pretrained model by specifying -a large(default) or -a small.

$ python3 u2net.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH -a small

When using ailia SDK 1.2.3 or earlier, you must use a lower accurate model by specifying --opset 10.

$ python3 u2net.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH --opset 10

Add the --composite option if you want to combine the input image with the calculated alpha value.

$ python3 u2net.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH --opset 11 --composite

Reference

U^2-Net: Going Deeper with Nested U-Structure for Salient Object Detection

Framework

PyTorch 1.1

Model Format

ONNX opset = 10

Netron

u2net.onnx.prototxt

u2netp.onnx.prototxt

u2net_opset11.onnx.prototxt

u2netp_opset11.onnx.prototxt