From Cradle to Cane: A Two-Pass Framework for High-Fidelity Lifespan Face Aging

NeurIPS 2025

If you find our work helpful, please consider giving us a ⭐!

📋Introduction

Abstract

Face aging has become a crucial task in computer vision, with applications ranging from entertainment to healthcare. However, existing methods struggle with achieving a realistic and seamless transformation across the entire lifespan, especially when handling large age gaps or extreme head poses. The core challenge lies in balancing age accuracy and identity preservation—what we refer to as the Age-ID trade-off. Most prior methods either prioritize age transformation at the expense of identity consistency or vice versa. In this work, we address this issue by proposing a two-pass face aging framework, named Cradle2Cane, based on few-step text-to-image (T2I) diffusion models. The first pass focuses on solving age accuracy by introducing an adaptive noise injection (AdaNI) mechanism. This mechanism is guided by including prompt descriptions of age and gender for the given person as the textual condition. Also, by adjusting the noise level, we can control the strength of aging while allowing more flexibility in transforming the face. However, identity preservation is weakly ensured here to facilitate stronger age transformations. In the second pass, we enhance identity preservation while maintaining age-specific features by conditioning the model on two identity-aware embeddings (SVR-ArcFace and Rotate-CLIP). This pass allows for denoising the transformed image from the first pass, ensuring stronger identity preservation without compromising the aging accuracy. Both passes are jointly trained in an end-to-end way. Extensive experiments on the CelebA-HQ test dataset, evaluated through Face++ and Qwen-VL protocols, show that Cradle2Cane outperforms existing face aging methods in age accuracy and identity consistency. Additionally, Cradle2Cane demonstrates superior robustness when applied to in-the-wild human face images, where prior methods often fail. This significantly broadens its applicability to more diverse and unconstrained real-world scenarios.

🛠️ Usage

1. Infer

git clone https://github.com/byliutao/Cradle2Cane

conda create --name cradle2cane python=3.10 -y
conda activate cradle2cane
pip install -r config/requirement.txt

# Download models
pip install -U huggingface_hub
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download --resume-download stabilityai/sdxl-turbo --local-dir models/sdxl-turbo/
huggingface-cli download --resume-download madebyollin/sdxl-vae-fp16-fix --local-dir models/sdxl-vae-fp16-fix/
huggingface-cli download --resume-download openai/clip-vit-large-patch14 --local-dir models/clip-vit-large-patch14/
huggingface-cli download --resume-download byliutao/Cradle2Cane --local-dir models/


# infer
python infer.py --input_path asserts/23_male.png

# in-the-wild image infer
python infer.py --input_path asserts/25_male.png --one_threshold

# infer with attribute change
python infer.py --input_path asserts/20_female.png --addition_prompt "yellow hair"

2. Training

Download the ffhq 512×512 dataset from the link and put the files to $REPOROOT/dataset.
Download the json directory from the link and put it under $REPOROOT/dataset/eval.
Download the ffhq-dataset-v2.json directory from the link and put it under $REPOROOT/dataset/eval.
The directory structure should look like:

$REPOROOT
|-- dataset
|   |-- ffhq512  # contains images :*.png
|   |-- json  # contains images :*.json
|   |-- ffhq-dataset-v2.json

# preprocess ffhq512 dataset
python -m lib.utils.ffhq_process

# check you config in train.sh first
bash config/train.sh

3. Evaluation

Download the celeba-200 dataset from the link and unzip the folder to $REPOROOT/dataset/eval.
Download the agedb-400 dataset from the link and unzip the folder to $REPOROOT/dataset/eval. (Follow Arc2Face, we use agebd dataset to do lpips evaluation)

The directory structure should look like:

$REPOROOT
|-- dataset
|   |-- eval  
|   |   |-- celeba-200  # contains images :*.jpg
|   |   |-- agedb-400  # contains images :*.jpg

# please set api key and secret first in eval.sh
bash config/eval.sh

4. Face Recognize Evaluation

install environment

git clone https://github.com/mk-minchul/AdaFace.git
cd Adaface

conda create --name adaface pytorch==1.8.0 torchvision==0.9.0 cudatoolkit=10.2 -c pytorch
conda activate adaface
conda install scikit-image matplotlib pandas scikit-learn 
pip install -r requirements.txt

Download dataset for face recognize

1. Download labeled faces_webface_112×112 dataset from link and unzip it to $REPOROOT/dataset/faces_webface_112x112
2. Download the faces_webface_112×112 dataset from link and unzip it to $REPOROOT/dataset/faces_webface_112x112.
3. run python convert.py --rec_path $REPOROOT/dataset/faces_webface_112x112 --make_image_files --make_validation_memfiles

The directory structure should look like:

$REPOROOT
|-- dataset  
|   |-- faces_webface_112x112  # contains subdir with imgs
|   |-- faces_webface_112x112_labeled  # contains images :*.jpg

Generate fake dataset and combine with real dataset

cd $REPOROOT
conda activate cradle2cane
bash config/gen_fake_fast.sh

Run training and eval

cd AdaFace
conda activate adaface 
bash ../config/run_ir50_ms1mv2.sh

🚀Results

24 Male	35 Female

📘Citation

If you find our paper or benchmark helpful for your research, please consider citing our paper and giving this repo a star ⭐. Thank you very much!

@inproceedings{
liu2025from,
title={From Cradle to Cane: A Two-Pass Framework for High-Fidelity Lifespan Face Aging},
author={Tao Liu and Dafeng Zhang and Gengchen Li and Shizhuo Liu and yongqi song and Senmao Li and Shiqi Yang and Boqian Li and Kai Wang and Yaxing Wang},
booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
year={2025},
url={https://openreview.net/forum?id=E1eVGJ5RYG}
}

License

Licensed under a Creative Commons Attribution-NonCommercial 4.0 International for Non-commercial use only. Any commercial use should get formal permission first.