Any2AnyTryon: AI-Powered Virtual Try-On Advances with Adaptive Position Embeddings

```html

Virtual Try-On Revolutionized with Any2AnyTryon

Virtual try-on of clothing (VTON) using image processing has made significant progress in recent years. The goal is to generate realistic representations of people in different clothes without them having to physically try on the clothing. However, previous methods faced challenges regarding generalization and quality due to the lack of paired data of clothing and people wearing them. This also limited the possibility of creating mask-free virtual try-ons.

Innovative approaches like Stable Garment and MMTryon use synthetic data to increase the amount of paired data. Despite these advances, existing methods often remain limited to specific try-on tasks and offer the user little flexibility. To improve the generalization and controllability of VTON generation, Any2AnyTryon was developed.

Adaptive Position Embedding as Key Technology

Any2AnyTryon enables the generation of try-on results based on various text instructions and images of clothing. This eliminates the need for masks, poses, or other conditions. A central component of Any2AnyTryon is adaptive position embedding. This technology allows the model to generate satisfactory images of people in outfits or even just clothing, based on input images of different sizes and categories. This significantly improves the generalization and controllability of VTON generation.

LAION-Garment: A New Dataset for Virtual Try-On

LAION-Garment, currently the largest known open-source dataset for virtual try-on, was created for training Any2AnyTryon. This dataset plays a crucial role in the performance of Any2AnyTryon and contributes to achieving more realistic and diverse results.

Versatile Applications and Benefits

Any2AnyTryon offers a flexible and controllable solution for the virtual try-on of clothing. By using text instructions, users can formulate specific requirements for the representation, such as the color or style of a garment. The adaptive position embedding ensures that the generated images are of high quality regardless of the size and category of the input images. This opens up new possibilities for online retail, the fashion industry, and also for private users.

Any2AnyTryon in the Context of AI Development

The development of Any2AnyTryon is in the context of the rapid progress in the field of artificial intelligence, particularly in the area of image generation and processing. Methods like Rotary Position Embedding and Conditional Image Generation contribute to creating increasingly realistic and detailed images. Any2AnyTryon utilizes these advancements to take the virtual try-on of clothing to a new level and push the boundaries of what is possible.

Outlook and Future Developments

Research in the field of virtual try-on is dynamic and promising. Future developments could include the integration of even more complex text instructions, the consideration of body shapes and sizes, and the expansion to other garments and accessories. Any2AnyTryon represents an important step towards realistic and user-friendly virtual try-on and has the potential to fundamentally change the way we buy and experience clothing online.

Bibliographie: Guo, H., Zeng, B., Song, Y., Zhang, W., Zhang, C., & Liu, J. (2025). Any2AnyTryon: Leveraging Adaptive Position Embeddings for Versatile Virtual Clothing Tasks. arXiv preprint arXiv:2501.15891. Su, J., Lu, Y., Pan, S., et al. (2021). RoFormer: Enhanced Transformer with Rotary Position Embedding. arXiv preprint arXiv:2104.09864. Karras, T., Aila, T., Laine, S., & Lehtinen, J. (2017). Progressive Growing of GANs for Improved Quality, Stability, and Variation. arXiv preprint arXiv:1710.10196. ```