Time-lapse image sequences offer visually compelling insights into dynamic processes that are too slow to observe in real time. However, playing a long time-lapse sequence back as a video often results in distracting flicker due to random effects, such as weather, as well as cyclic effects, such as the day-night cycle. We introduce the problem of disentangling time-lapse sequences in a way that allows separate, after-the-fact control of overall trends, cyclic effects, and random effects in the images, and describe a technique based on data-driven generative models that achieves this goal. This enables us to "re-render" the sequences in ways that would not be possible with the input images alone. For example, we can stabilize a long sequence to focus on plant growth over many months, under selectable, consistent weather. Our approach is based on Generative Adversarial Networks (GAN) that are conditioned with the time coordinate of the time-lapse sequence. Our architecture and training procedure are designed so that the networks learn to model random variations, such as weather, using the GAN's latent space, and to disentangle overall trends and cyclic variations by feeding the conditioning time label to the model using Fourier features with specific frequencies. We show that our models are robust to defects in the training data, enabling us to amend some of the practical difficulties in capturing long time-lapse sequences, such as temporary occlusions, uneven frame spacing, and missing frames.

We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation.

This paper describes a simple technique to analyze Generative Adversarial Networks (GANs) and create interpretable controls for image synthesis, such as change of viewpoint, aging, lighting, and time of day. We identify important latent directions based on Principal Component Analysis (PCA) applied either in latent space or feature space. Then, we show that a large number of interpretable controls can be defined by layer-wise perturbation along the principal directions. Moreover, we show that BigGAN can be controlled with layer-wise inputs in a StyleGAN-like manner. We show results on different GANs trained on various datasets, and demonstrate good qualitative matches to edit directions found through earlier supervised approaches.

It has been shown that rendering in the gradient domain, i.e., estimating finite difference gradients of image intensity using correlated samples, and combining them with direct estimates of pixel intensities by solving a screened Poisson problem, often offers fundamental benefits over merely sampling pixel intensities. The reasons can be traced to the frequency content of the light transport integrand and its interplay with the gradient operator. However, while they often yield state of the art performance among algorithms that are based on Monte Carlo sampling alone, gradient-domain rendering algorithms have, until now, not generally been competitive with techniques that combine Monte Carlo sampling with post-hoc noise removal using sophisticated non-linear filtering. Drawing on the power of modern convolutional neural networks, we propose a novel reconstruction method for gradient-domain rendering. Our technique replaces the screened Poisson solver of previous gradient-domain techniques with a novel dense variant of the U-Net autoencoder, additionally taking auxiliary feature buffers as inputs. We optimize our network to minimize a perceptual image distance metric calibrated to the human visual system. Our results significantly improve the quality obtained from gradient-domain path tracing, allowing it to overtake state-of-the-art comparison techniques that denoise traditional Monte Carlo samplings. In particular, we observe that the correlated gradient samples --- that offer information about the smoothness of the integrand unavailable in standard Monte Carlo sampling --- notably improve image quality compared to an equally powerful neural model that does not make use of gradient samples.