sam

Age Transformation Using a Style-Based Regression Model

Input

(Image from https://github.com/yuval-alaluf/SAM/blob/master/notebooks/images/866.jpg)

Shape : (1, 3, 1024, 1024)

Face alignment and reshaped to : (1, 3, 256, 256)

Output

* Note: From left to right: original (face aligned) image, 10-, 30-, 50-, 70-, and 90-year-old-individual images.

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 sam.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 sam.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH 

By specifying the -age option, you can choose the individual's age(s) (between 0 and 100) to be generated (default '10,30,50,70,90').

$ python3 sam.py -age '0,25,50,75,100' 

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 sam.py --video VIDEO_PATH 

By adding the --use_dlib option, you can use original version of face alignment.

Reference

• SAM

• PSGAN (face alignment without dlib)

Framework

Pytorch 1.10.0

Python 3.6.7+

Model Format

ONNX opset=11

Netron

encoder.onnx.prototxt

pretrained-encoder.onnx.prototxt

decoder.onnx.prototxt