190 papers with code • 15 benchmarks • 14 datasets
( Various Video Generation Tasks. Gif credit: MaGViT )
Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible.
Additionally, we propose a first set of metrics to quantitatively evaluate the accuracy as well as the perceptual quality of the temporal evolution.
In this paper, we propose a generative model, Temporal Generative Adversarial Nets (TGAN), which can learn a semantic representation of unlabeled videos, and is capable of generating videos.
However, learning to predict raw future observations, such as frames in a video, is exceedingly challenging -- the ambiguous nature of the problem can cause a naively designed model to average together possible futures into a single, blurry prediction.
We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes.
A core challenge for an agent learning to interact with the world is to predict how its actions affect objects in its environment.