Animesh Karnewar

Animesh Karnewar, Andrea Vedaldi, David Novotny et al.

Diffusion models have emerged as the best approach for generative modeling of 2D images. Part of their success is due to the possibility of training them on millions if not billions of images with a stable learning objective. However, extending these models to 3D remains difficult for two reasons. First, finding a large quantity of 3D training data is much more complex than for 2D images. Second, while it is conceptually trivial to extend the models to operate on 3D rather than 2D grids, the associated cubic growth in memory and compute complexity makes this infeasible. We address the first challenge by introducing a new diffusion setup that can be trained, end-to-end, with only posed 2D images for supervision; and the second challenge by proposing an image formation model that decouples model memory from spatial memory. We evaluate our method on real-world data, using the CO3D dataset which has not been used to train 3D generative models before. We show that our diffusion models are scalable, train robustly, and are competitive in terms of sample quality and fidelity to existing approaches for 3D generative modeling.

Animesh Karnewar, Tobias Ritschel, Oliver Wang et al.

In many recent works, multi-layer perceptions (MLPs) have been shown to be suitable for modeling complex spatially-varying functions including images and 3D scenes. Although the MLPs are able to represent complex scenes with unprecedented quality and memory footprint, this expressive power of the MLPs, however, comes at the cost of long training and inference times. On the other hand, bilinear/trilinear interpolation on regular grid based representations can give fast training and inference times, but cannot match the quality of MLPs without requiring significant additional memory. Hence, in this work, we investigate what is the smallest change to grid-based representations that allows for retaining the high fidelity result of MLPs while enabling fast reconstruction and rendering times. We introduce a surprisingly simple change that achieves this task -- simply allowing a fixed non-linearity (ReLU) on interpolated grid values. When combined with coarse to-fine optimization, we show that such an approach becomes competitive with the state-of-the-art. We report results on radiance fields, and occupancy fields, and compare against multiple existing alternatives. Code and data for the paper are available at https://geometry.cs.ucl.ac.uk/projects/2022/relu_fields.

Animesh Karnewar, Niloy J. Mitra, Andrea Vedaldi et al.

Diffusion-based image generators can now produce high-quality and diverse samples, but their success has yet to fully translate to 3D generation: existing diffusion methods can either generate low-resolution but 3D consistent outputs, or detailed 2D views of 3D objects but with potential structural defects and lacking view consistency or realism. We present HoloFusion, a method that combines the best of these approaches to produce high-fidelity, plausible, and diverse 3D samples while learning from a collection of multi-view 2D images only. The method first generates coarse 3D samples using a variant of the recently proposed HoloDiffusion generator. Then, it independently renders and upsamples a large number of views of the coarse 3D model, super-resolves them to add detail, and distills those into a single, high-fidelity implicit 3D representation, which also ensures view consistency of the final renders. The super-resolution network is trained as an integral part of HoloFusion, end-to-end, and the final distillation uses a new sampling scheme to capture the space of super-resolved signals. We compare our method against existing baselines, including DreamFusion, Get3D, EG3D, and HoloDiffusion, and achieve, to the best of our knowledge, the most realistic results on the challenging CO3Dv2 dataset.

Animesh Karnewar, Oliver Wang, Tobias Ritschel et al.

We introduce 3inGAN, an unconditional 3D generative model trained from 2D images of a single self-similar 3D scene. Such a model can be used to produce 3D "remixes" of a given scene, by mapping spatial latent codes into a 3D volumetric representation, which can subsequently be rendered from arbitrary views using physically based volume rendering. By construction, the generated scenes remain view-consistent across arbitrary camera configurations, without any flickering or spatio-temporal artifacts. During training, we employ a combination of 2D, obtained through differentiable volume tracing, and 3D Generative Adversarial Network (GAN) losses, across multiple scales, enforcing realism on both its 3D structure and the 2D renderings. We show results on semi-stochastic scenes of varying scale and complexity, obtained from real and synthetic sources. We demonstrate, for the first time, the feasibility of learning plausible view-consistent 3D scene variations from a single exemplar scene and provide qualitative and quantitative comparisons against recent related methods.

Denis Korzhenkov, Adil Karjauv, Animesh Karnewar et al.

Recently proposed pyramidal models decompose the conventional forward and backward diffusion processes into multiple stages operating at varying resolutions. These models handle inputs with higher noise levels at lower resolutions, while less noisy inputs are processed at higher resolutions. This hierarchical approach significantly reduces the computational cost of inference in multi-step denoising models. However, existing open-source pyramidal video models have been trained from scratch and tend to underperform compared to state-of-the-art systems in terms of visual plausibility. In this work, we present a pipeline that converts a pretrained diffusion model into a pyramidal one through low-cost finetuning, achieving this transformation without degradation in quality of output videos. Furthermore, we investigate and compare various strategies for step distillation within pyramidal models, aiming to further enhance the inference efficiency. Our results are available at https://qualcomm-ai-research.github.io/PyramidalWan.

Animesh Karnewar, Denis Korzhenkov, Ioannis Lelekas et al.

We introduce Neodragon, a text-to-video system capable of generating 2s (49 frames @24 fps) videos at the 640x1024 resolution directly on a Qualcomm Hexagon NPU in a record 6.7s (7 FPS). Differing from existing transformer-based offline text-to-video generation models, Neodragon is the first to have been specifically optimised for mobile hardware to achieve efficient and high-fidelity video synthesis. We achieve this through four key technical contributions: (1) Replacing the original large 4.762B T5xxl Text-Encoder with a much smaller 0.2B DT5 (DistilT5) with minimal quality loss, enabled through a novel Text-Encoder Distillation procedure. (2) Proposing an Asymmetric Decoder Distillation approach allowing us to replace the native codec-latent-VAE decoder with a more efficient one, without disturbing the generative latent-space of the generation pipeline. (3) Pruning of MMDiT blocks within the denoiser backbone based on their relative importance, with recovery of original performance through a two-stage distillation process. (4) Reducing the NFE (Neural Functional Evaluation) requirement of the denoiser by performing step distillation using DMD adapted for pyramidal flow-matching, thereby substantially accelerating video generation. When paired with an optimised SSD1B first-frame image generator and QuickSRNet for 2x super-resolution, our end-to-end Neodragon system becomes a highly parameter (4.945B full model), memory (3.5GB peak RAM usage), and runtime (6.7s E2E latency) efficient mobile-friendly model, while achieving a VBench total score of 81.61. By enabling low-cost, private, and on-device text-to-video synthesis, Neodragon democratizes AI-based video content creation, empowering creators to generate high-quality videos without reliance on cloud services. Code and model will be made publicly available at our website: https://qualcomm-ai-research.github.io/neodragon

Raphael Bensadoun, Tom Monnier, Yanir Kleiman et al.

We introduce Meta 3D Gen (3DGen), a new state-of-the-art, fast pipeline for text-to-3D asset generation. 3DGen offers 3D asset creation with high prompt fidelity and high-quality 3D shapes and textures in under a minute. It supports physically-based rendering (PBR), necessary for 3D asset relighting in real-world applications. Additionally, 3DGen supports generative retexturing of previously generated (or artist-created) 3D shapes using additional textual inputs provided by the user. 3DGen integrates key technical components, Meta 3D AssetGen and Meta 3D TextureGen, that we developed for text-to-3D and text-to-texture generation, respectively. By combining their strengths, 3DGen represents 3D objects simultaneously in three ways: in view space, in volumetric space, and in UV (or texture) space. The integration of these two techniques achieves a win rate of 68% with respect to the single-stage model. We compare 3DGen to numerous industry baselines, and show that it outperforms them in terms of prompt fidelity and visual quality for complex textual prompts, while being significantly faster.

Animesh Karnewar, Roman Shapovalov, Tom Monnier et al. · meta-ai

Encoding information from 2D views of an object into a 3D representation is crucial for generalized 3D feature extraction. Such features can then enable 3D reconstruction, 3D generation, and other applications. We propose GOEmbed (Gradient Origin Embeddings) that encodes input 2D images into any 3D representation, without requiring a pre-trained image feature extractor; unlike typical prior approaches in which input images are either encoded using 2D features extracted from large pre-trained models, or customized features are designed to handle different 3D representations; or worse, encoders may not yet be available for specialized 3D neural representations such as MLPs and hash-grids. We extensively evaluate our proposed GOEmbed under different experimental settings on the OmniObject3D benchmark. First, we evaluate how well the mechanism compares against prior encoding mechanisms on multiple 3D representations using an illustrative experiment called Plenoptic-Encoding. Second, the efficacy of the GOEmbed mechanism is further demonstrated by achieving a new SOTA FID of 22.12 on the OmniObject3D generation task using a combination of GOEmbed and DFM (Diffusion with Forward Models), which we call GOEmbedFusion. Finally, we evaluate how the GOEmbed mechanism bolsters sparse-view 3D reconstruction pipelines.

Phong Nguyen-Ha, Animesh Karnewar, Lam Huynh et al.

We propose a new cascaded architecture for novel view synthesis, called RGBD-Net, which consists of two core components: a hierarchical depth regression network and a depth-aware generator network. The former one predicts depth maps of the target views by using adaptive depth scaling, while the latter one leverages the predicted depths and renders spatially and temporally consistent target images. In the experimental evaluation on standard datasets, RGBD-Net not only outperforms the state-of-the-art by a clear margin, but it also generalizes well to new scenes without per-scene optimization. Moreover, we show that RGBD-Net can be optionally trained without depth supervision while still retaining high-quality rendering. Thanks to the depth regression network, RGBD-Net can be also used for creating dense 3D point clouds that are more accurate than those produced by some state-of-the-art multi-view stereo methods.

Animesh Karnewar, Oliver Wang

While Generative Adversarial Networks (GANs) have seen huge successes in image synthesis tasks, they are notoriously difficult to adapt to different datasets, in part due to instability during training and sensitivity to hyperparameters. One commonly accepted reason for this instability is that gradients passing from the discriminator to the generator become uninformative when there isn't enough overlap in the supports of the real and fake distributions. In this work, we propose the Multi-Scale Gradient Generative Adversarial Network (MSG-GAN), a simple but effective technique for addressing this by allowing the flow of gradients from the discriminator to the generator at multiple scales. This technique provides a stable approach for high resolution image synthesis, and serves as an alternative to the commonly used progressive growing technique. We show that MSG-GAN converges stably on a variety of image datasets of different sizes, resolutions and domains, as well as different types of loss functions and architectures, all with the same set of fixed hyperparameters. When compared to state-of-the-art GANs, our approach matches or exceeds the performance in most of the cases we tried.