论文复现：Only a Matter of Style: Age Transformation Using a Style-Based Regression Model

---

English|简体中文

• paddle-SAM
  • 1 Introduction
  • 2 Result
  • 3 Datasets
  • 4 Environment
  • 5 Pretrained models
  • 6 Quick start
    • Inference
    • Train
    • Others
  • 7 Code structure
    • Structure
    • Parameter description
  • 8 Model information

1 Introduction

---

This protest reproduces SAM based on paddlepaddle framework.SAM is an image-to-imagetranslation method that learns to directly encode real facial images into the latent space of a pre-trained unconditional GAN subject to a given aging shift.

Paper

• [1] Y Alaluf, Patashnik O , Cohen-Or D . Only a Matter of Style: Age Transformation Using a Style-Based Regression Model[J]. 2021.

Reference project

• https://github.com/yuval-alaluf/SAM

Project on Ai Studio

• notebook

https://aistudio.baidu.com/aistudio/projectdetail/2331297

2 Result

• The current presented is the result of the model that is saved 24,000 steps. According to the author's results, the results are running 60000 steps. The picture from left to right is: Enter the picture, the model is 0 years old, 10 years old, 20 years old, 30 Years, 40 years old, 60 years old, 70 years old, 80 years old, 90 years old, 100 years old.

Visual comparison

模型	图片
Pytorch
Paddle
Pytorch
Paddle
Pytorch
Paddle

3 Datasets

• Training： FFHQ-1024. saved in SAM/data/FFHQ/.

• Testing：CelebA-HQ.saved inSAM/data/CelebA_test/.

4 Environment

Hardware：GPU、CPU

Framework：PaddlePaddle >=2.0.0

5 Pretrained models

Pretrained models saved inpretrained_models/.

Pretrained models	Description
FFHQ StyleGAN(stylegan2-ffhq-config-f.pdparams)	StyleGAN trained with the FFHQ dataset fromrosinality ，output size:1024x1024
IR-SE50 Model(model_ir_se50.pdparams)	IR_SE model (TreB1eN)trained for computering ID loss.
CurricularFace Backbone(CurricularFace_Backbone.paparams)	Pretrained CurricularFace model(HuangYG123)evaled Similarity
AlexNet(alexnet.pdparams和lin_alex.pdparams)	computered lpips loss
StyleGAN Inversion(psp_ffhq_inverse.pdparams)	pSp trained with the FFHQ dataset for StyleGAN inversion.

Baidu driver：https://pan.baidu.com/s/1G-Ffs8-y93R0ZlD9mEU6Eg password：m3nb

6 Quick start

# clone this repo
git clone [https://github.com/771979972/paddle-SAM.git]
cd work

Inference

python SAM/scripts/inference_side_by_side.py 
--exp_dir=exp/test 
--checkpoint_path=SAM/pretrained_models/sam_ffhq_aging.pdparams 
--data_path=SAM/data/CelebA_test 
--test_batch_size=4 
--test_workers=0 
--target_age=0,10,20,30,40,50,60,70,80,90,100

Configuration Environment

!pip install --upgrade matplotlib
python SAM/scripts/compile_ranger.py

Train

python SAM/scripts/train.py /
--dataset_type=ffhq_encode /
--exp_dir=exp/test /
--workers=0 /
--batch_size=8/
--test_batch_size=8 /
--test_workers=0 /
--val_interval=2500 /
--save_interval=5000 /
--encoder_type=GradualStyleEncoder/
--start_from_latent_avg /
--lpips_lambda=0.8 \--l2_lambda=1 /
--id_lambda=0.1 /
--optim_name=ranger

Others

LPIPS

python SAM/scripts/calc_losses_on_images.py /
--mode lpips /
--data_path=SAM/inference/inference_results /
--gt_path=SAM/data/CelebA_test/

MSE

python SAM/scripts/calc_losses_on_images.py /
--mode l2 /
--data_path=SAM/inference/inference_results /
--gt_path=SAM/data/CelebA_test/

Similarity

python SAM/scripts/calc_id_loss_parallel.py /
--data_path=SAM/inference/inference_results /
--gt_path=SAM/data/CelebA_test/

7 Code structure

Structure

├─config          # 配置
├─data            #数据集加载
   ├─CelebA_test  # 测试数据图像
├─models          # 模型
    ├─encoders    # 编码器
    ├─loss        # 损失函数
    ├─utils       # 编译算子
├─scripts         #算法执行
    trian         #训练
    inference     #测试
    inference_side_by_side    #测试
    reference_guided_inference    #测试
├─utils           # 工具代码
│  README.md      #英文readme
│  README_cn.md   #中文readme

Parameter description

Parameter	Default
config	None
dataset_type	ffhq_aging
exp_dir	exp/test
workers	0
test_workers	0
batch_size	6
test_batch_size	6
start_from_encoded_w_plus	store-true
use_weighted_id_loss	store-true
id_lambda	0.1
lpips_lambda	0.1
lpips_lambda_aging	0.1
lpips_lambda_crop	0.6
l2_lambda	0.25
l2_lambda_aging	0，25
l2_lambda_crop	1
w_norm_lambda	0，005
aging_lambda	5
cycle_lambda	1
input_nc	4
target_age	uniform_random

8 Model information

The overall information of the model is as follows:

Information	Descriptions
Version	Paddle 2.1.2
Application	Image Generation
Hardware	GPU / CPU

License

#encoding=utf8
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.

# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at

#     http://www.apache.org/licenses/LICENSE-2.0

# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.