J. P. Lewis

Yuval Atzmon, Maciej Bala, Yogesh Balaji et al. · nvidia

We introduce Edify Image, a family of diffusion models capable of generating photorealistic image content with pixel-perfect accuracy. Edify Image utilizes cascaded pixel-space diffusion models trained using a novel Laplacian diffusion process, in which image signals at different frequency bands are attenuated at varying rates. Edify Image supports a wide range of applications, including text-to-image synthesis, 4K upsampling, ControlNets, 360 HDR panorama generation, and finetuning for image customization.

Wan-Duo Kurt Ma, J. P. Lewis, Avisek Lahiri et al.

Text-guided diffusion models such as DALLE-2, Imagen, eDiff-I, and Stable Diffusion are able to generate an effectively endless variety of images given only a short text prompt describing the desired image content. In many cases the images are of very high quality. However, these models often struggle to compose scenes containing several key objects such as characters in specified positional relationships. The missing capability to ``direct'' the placement of characters and objects both within and across images is crucial in storytelling, as recognized in the literature on film and animation theory. In this work, we take a particularly straightforward approach to providing the needed direction. Drawing on the observation that the cross-attention maps for prompt words reflect the spatial layout of objects denoted by those words, we introduce an optimization objective that produces ``activation'' at desired positions in these cross-attention maps. The resulting approach is a step toward generalizing the applicability of text-guided diffusion models beyond single images to collections of related images, as in storybooks. Directed Diffusion provides easy high-level positional control over multiple objects, while making use of an existing pre-trained model and maintaining a coherent blend between the positioned objects and the background. Moreover, it requires only a few lines to implement.

Wan-Duo Kurt Ma, J. P. Lewis, W. Bastiaan Kleijn

Within recent approaches to text-to-video (T2V) generation, achieving controllability in the synthesized video is often a challenge. Typically, this issue is addressed by providing low-level per-frame guidance in the form of edge maps, depth maps, or an existing video to be altered. However, the process of obtaining such guidance can be labor-intensive. This paper focuses on enhancing controllability in video synthesis by employing straightforward bounding boxes to guide the subject in various ways, all without the need for neural network training, finetuning, optimization at inference time, or the use of pre-existing videos. Our algorithm, TrailBlazer, is constructed upon a pre-trained (T2V) model, and easy to implement. The subject is directed by a bounding box through the proposed spatial and temporal attention map editing. Moreover, we introduce the concept of keyframing, allowing the subject trajectory and overall appearance to be guided by both a moving bounding box and corresponding prompts, without the need to provide a detailed mask. The method is efficient, with negligible additional computation relative to the underlying pre-trained model. Despite the simplicity of the bounding box guidance, the resulting motion is surprisingly natural, with emergent effects including perspective and movement toward the virtual camera as the box size increases.

Wan-Duo Kurt Ma, J. P. Lewis, W. Bastiaan Kleijn

We introduce the HSIC (Hilbert-Schmidt independence criterion) bottleneck for training deep neural networks. The HSIC bottleneck is an alternative to the conventional cross-entropy loss and backpropagation that has a number of distinct advantages. It mitigates exploding and vanishing gradients, resulting in the ability to learn very deep networks without skip connections. There is no requirement for symmetric feedback or update locking. We find that the HSIC bottleneck provides performance on MNIST/FashionMNIST/CIFAR10 classification comparable to backpropagation with a cross-entropy target, even when the system is not encouraged to make the output resemble the classification labels. Appending a single layer trained with SGD (without backpropagation) to reformat the information further improves performance.