The Google team announced the creation of Transframer, a new program that can generate short videos based on individual images.
This technology will complement traditional rendering solutions by allowing developers to create virtual environments based on machine learning capabilities.
The program was named after another artificial intelligence model known as Transformer. Transformer is a new neural network architecture capable of generating text by simulating and comparing other words in a sentence. The model was created in 2017 and has since been used in standard deep learning frameworks such as TensorFlow and PyTorch.
Just as Transformer uses language to predict potential results, Transframer uses an image combined with a query description to create short videos. Transframer uses the Google DeepMind artificial intelligence platform to work.
The technology analyzes a single image to extract key fragments and create additional images. The contextual images are then used to further predict how the image will look from different angles.
In addition to being able to generate video from a very limited data set, Transframer also performs well in other video-related tasks and tests: semantic segmentation, image classification, and optical flow prediction.
Earlier it was reported that Google is updating Search to better deal with clickbait and give preference to real reviews.
Source: tech spot