Using A4 paper as the screen to "play" Hayao Miyazaki's animation, there are no flaws in any random shaking!

The content of this article comes from the internet. If it does not match the actual situation or there is infringement, please contact to delete it

The content of this article comes from the internet. If it does not match the actual situation or there is infringement, please contact to delete it.

In a magical experiment, a group of researchers from the Chinese University of Hong Kong, Zhejiang University, and Nvidia demonstrated amazing AI synthetic image technology. They not only successfully moved a plate of hot dogs and plates in the video, but also moved Hayao Miyazaki's "Dragon Cat" onto a blank sheet of paper, bending and moving it freely without affecting the playback.

Behind this seemingly incredible scene, a new framework called NeuralMarker is actually used. This framework achieves a silky and realistic visual effect by capturing the dense correspondence between markers and reference images. Unlike previous models, NeuralMarker performs well in handling geometric and color brightness changes. For geometric changes, they developed the FlyingMarkers tool, which successfully captured various deformations. For changes in color brightness, they introduced the concept of symmetrical polar distance loss and combined it with motion structure technology to produce excellent images even in dim light.

In addition, researchers have proposed a new marker correspondence evaluation method: DVL-Markers, which is used to evaluate the correspondence between pixel blocks in real captured images and the markers contained in the model. A comprehensive evaluation was conducted from geometric deformation, camera perspective, to lighting. After testing, the image quality evaluation indicators SSIM and PSNR of this AI model perform better than similar models, and the task success rate has reached an astonishing 100%.

However, although NeuralMarker performs well, there are also some limitations, such as when there are obstacles in front of the image, this AI appears less intelligent. However, the researchers stated that they will randomly add obstacles to the training dataset to further enhance the intelligence level of this AI.

This study is not only about simple image processing, but also takes AI technology to a new level. Researchers combined NeuralMarker with NeRF and directly added a realistic three-dimensional object to the two-dimensional image, as if integrating virtual and reality.

However, researchers also admit that NeuralMarker has some limitations, especially when there are obstacles in front of the image, the intelligent performance of this AI will be affected. In order to overcome this challenge, they plan to randomly introduce obstacles in the training dataset to make this AI perform well in more complex situations.

This groundbreaking research achievement is not only of great significance at the technical level, but also opens up new directions for future technological development. In the real world, we may be able to see more similar application scenarios that integrate the virtual world with the physical environment, bringing unprecedented visual experiences to people.

Overall, this study demonstrates the enormous potential of AI synthetic image technology and presents us with a completely new possibility. With the continuous development of technology, we can expect to see more innovation based on artificial intelligence, bringing more surprises to our lives. This also makes us wonder what the future technological world will become, and what role will we play in it?

With the publication of this breakthrough research, people are eagerly anticipating the future of AI in the field of visual technology. Perhaps in the near future, we will be able to see more applications based on this type of technology in our daily lives, whether in entertainment, education, or other fields.

This study also makes us realize that, driven by technology, every aspect of our lives will undergo profound changes. The boundary between virtual and reality will become increasingly blurred, and we will be able to enjoy unprecedented immersive experiences.

However, we should also recognize that behind this breakthrough lies the long-term hard work and continuous innovation of countless researchers. They achieved such remarkable results through the exploration and practice of technology.

With the continuous development and improvement of NeuralMarker technology, we can look forward to seeing more surprises based on this technology in the future. Perhaps we will be able to watch more dynamic images on paper, or witness similar breakthroughs in other fields.

In the end, this research achievement opened a door to the future of technology for us. It is not only a triumph of technology, but also a crystallization of human wisdom, leading us towards an unprecedented era of visual experience. We can't wait to see what surprises and changes this technology will bring us in the future.

The above content and information are all sourced from the internet, relevant data, and theoretical research is based on online materials. It does not mean that the author of this article agrees with the laws, rules, opinions, behaviors, and is responsible for the authenticity of the relevant information in the article. We are not responsible for any issues arising from the above or related matters, and the author of this article does not assume any direct or indirect legal responsibility.

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])