Google's artificial intelligence technology "Transframer" can create short videos based on a picture

As technology evolves, researchers continue to find new ways to leverage artificial intelligence and machine learning capabilities. Earlier this week, Google scientists announced the creation of Transframer, a new framework for generating short videos from a single image input. This new technology could one day enhance traditional rendering solutions and enable developers to create virtual environments based on machine learning capabilities.

The name (and in some ways the concept) of this new framework is a nod to Transformer, another AI-based model. Originally launched in 2017, Transformer is a novel neural network architecture that has the ability to generate text by modeling and comparing other words in a sentence. The model has since been incorporated into standard deep learning frameworks such as TensorFlow and PyTorch.

It is reported that Transframer uses background images with similar attributes, combined with query annotations, to create short videos. Although no geometric data is provided in the raw image input, the resulting video moves around the target image and visualizes the accurate perspective.

​​​​

The new technology was demonstrated using Google’s DeepMind artificial intelligence platform, which features analytics A single photo background image is used to capture key image data and generate additional images. During this analysis, the system determines the frame of the image, which in turn helps the system predict the image's surroundings.

Context images are then used to further predict how the picture will appear from different angles. Prediction models the probability of additional image frames based on the data, annotations, and any other information in the contextual frame.

This framework marks a huge advancement in video technology by providing the ability to generate reasonably accurate videos based on very limited data sets. The Transframer task also shows promising results on other video-related tasks and benchmarks, such as semantic segmentation, image classification, and optical flow prediction.

Could have potentially huge impact on video-based industries such as game development. Current game development environments rely on core rendering technologies such as shading, texture mapping, depth of field, and ray tracing. Technologies like Transframer have the potential to offer developers a new development path by using artificial intelligence and machine learning to build their environments and at the same time reduce the time, resources and effort required to create them.

The above is the detailed content of Google's artificial intelligence technology "Transframer" can create short videos based on a picture. For more information, please follow other related articles on the PHP Chinese website!