MyTimeMachine - AI-powered personalized facial age transformation technology, enabling age transitions spanning 20 to 40 years.

What is MyTimeMachine?

MyTimeMachine (MyTM) is an advanced personalized facial aging technology that trains a globally aging model using an adapter network based on approximately 50 individual photographs spanning 20 to 40 years. MyTimeMachine achieves high-quality age regression and aging progression effects while preserving individual identity features. MyTimeMachine can be extended to the video domain to generate aging effects with high identity preservation and temporal consistency, surpassing existing technologies.

Main functions of MyTimeMachine

• Personalized age conversionAge transformation of faces based on a collection of personal photos, including de-aging and aging.
• Identity preservationWhile converting the age, the facial features of the person are preserved to ensure that the converted image matches the facial features of the original person.
• High-quality image generationBased on advanced deep learning technology, it generates high-resolution, realistic facial images.
• Video extensionIn addition to processing static images, it extends to videos, enabling dynamic changes in the age of people in videos while maintaining consistency over time.
• AdaptabilityIt adapts to different age ranges, performs well within the age range covered by the training data, and optimizes its performance when extrapolated to unseen age ranges.

The technical principles of MyTimeMachine

• Adapter NetworkAn adapter network is introduced that combines personalized aging features and global aging features to generate aged images based on StyleGAN2.
• loss functionFor personalized adapter networks, three loss functions are introduced:
  • Personalized aging lossEnsure that the aged images are similar in identity features to reference images of similar age in the personal photo collection.
  • Extrapolation regularizationTo control the aging effects beyond the training age range, a global prior is used.
  • Adaptive w-norm regularization: Solve the inversion-editability tradeoff of StyleGAN, ensuring that aging changes in shape and texture are performed while maintaining identity.
• Global Aging PriorBased on a pre-trained global aging model, the model learns the aging patterns of the general population.
• Personal photo collectionUsers need to provide approximately 50 personal photos, spanning a certain age range, to be used in training the adapter network to learn personalized aging characteristics.
• Extend to videoBased on face swapping technology, personalized aging effects are applied to videos to generate aging videos that are consistent in time.

MyTimeMachine project address

• Project official website:mytimemachine.github.io
• arXiv technical paper:https://arxiv.org/pdf/2411.14521

Application scenarios of MyTimeMachine

• Film and television productionIn movies and TV series, creating effects that change the age of characters, such as making actors look younger or older, to suit the needs of the plot.
• Advertising and Entertainment: Show the effects of a product over time in an advertisement, such as skincare or health products.
• Forensic medicine and criminal investigationIt helps law enforcement agencies identify and track suspects or missing persons who have not been seen for many years through aging images.
• History repeats itself: To recreate the image of historical figures at different ages, for use in education or historical documentaries.
• Personal entertainmentUsers can imagine what they will look like at a certain age in the future, for entertainment or special commemoration.