Neural Radiance Field (NeRF) is a representation for 3D reconstruction from multi-view images. Despite some recent work showing preliminary success in editing a reconstructed NeRF with diffusion prior, they remain struggling to synthesize reasonable geometry in completely uncovered regions. One major reason is the high diversity of synthetic contents from the diffusion model, which hinders the radiance field from converging to a crisp and deterministic geometry. Moreover, applying latent diffusion models on real data often yields a textural shift incoherent to the image condition due to auto-encoding errors. These two problems are further reinforced with the use of pixel-distance losses. To address these issues, we propose tempering the diffusion model's stochasticity with per-scene customization and mitigating the textural shift with masked adversarial training. During the analyses, we also found the commonly used pixel and perceptual losses are harmful in the NeRF inpainting task. Through rigorous experiments, our framework yields state-of-the-art NeRF inpainting results on various real-world scenes.

Toward unlocking the potential of generative models in immersive 4D experiences, we introduce Virtual Pet, a novel pipeline to model realistic and diverse motions for target animal species within a 3D environment. To circumvent the limited availability of 3D motion data aligned with environmental geometry, we leverage monocular internet videos and extract deformable NeRF representations for the foreground and static NeRF representations for the background. For this, we develop a reconstruction strategy, encompassing species-level shared template learning and per-video fine-tuning. Utilizing the reconstructed data, we then train a conditional 3D motion model to learn the trajectory and articulation of foreground animals in the context of 3D backgrounds. We showcase the efficacy of our pipeline with comprehensive qualitative and quantitative evaluations using cat videos. We also demonstrate versatility across unseen cats and indoor environments, producing temporally coherent 4D outputs for enriched virtual experiences.

Articulated 3D reconstruction has valuable applications in various domains, yet it remains costly and demands intensive work from domain experts. Recent advancements in template-free learning methods show promising results with monocular videos. Nevertheless, these approaches necessitate a comprehensive coverage of all viewpoints of the subject in the input video, thus limiting their applicability to casually captured videos from online sources. In this work, we study articulated 3D shape reconstruction from a single and casually captured internet video, where the subject's view coverage is incomplete. We propose DreaMo that jointly performs shape reconstruction while solving the challenging low-coverage regions with view-conditioned diffusion prior and several tailored regularizations. In addition, we introduce a skeleton generation strategy to create human-interpretable skeletons from the learned neural bones and skinning weights. We conduct our study on a self-collected internet video collection characterized by incomplete view coverage. DreaMo shows promising quality in novel-view rendering, detailed articulated shape reconstruction, and skeleton generation. Extensive qualitative and quantitative studies validate the efficacy of each proposed component, and show existing methods are unable to solve correct geometry due to the incomplete view coverage.

Recent advancements in diffusion models have greatly improved the quality and diversity of synthesized content. To harness the expressive power of diffusion models, researchers have explored various controllable mechanisms that allow users to intuitively guide the content synthesis process. Although the latest efforts have primarily focused on video synthesis, there has been a lack of effective methods for controlling and describing desired content and motion. In response to this gap, we introduce MCDiff, a conditional diffusion model that generates a video from a starting image frame and a set of strokes, which allow users to specify the intended content and dynamics for synthesis. To tackle the ambiguity of sparse motion inputs and achieve better synthesis quality, MCDiff first utilizes a flow completion model to predict the dense video motion based on the semantic understanding of the video frame and the sparse motion control. Then, the diffusion model synthesizes high-quality future frames to form the output video. We qualitatively and quantitatively show that MCDiff achieves the state-the-of-art visual quality in stroke-guided controllable video synthesis. Additional experiments on MPII Human Pose further exhibit the capability of our model on diverse content and motion synthesis.

Toward infinite-scale 3D city synthesis, we propose a novel framework, InfiniCity, which constructs and renders an unconstrainedly large and 3D-grounded environment from random noises. InfiniCity decomposes the seemingly impractical task into three feasible modules, taking advantage of both 2D and 3D data. First, an infinite-pixel image synthesis module generates arbitrary-scale 2D maps from the bird's-eye view. Next, an octree-based voxel completion module lifts the generated 2D map to 3D octrees. Finally, a voxel-based neural rendering module texturizes the voxels and renders 2D images. InfiniCity can thus synthesize arbitrary-scale and traversable 3D city environments, and allow flexible and interactive editing from users. We quantitatively and qualitatively demonstrate the efficacy of the proposed framework.

Recent studies have shown that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring and visualizing the encoded positional information. We formally define the encoded information as Position-information Pattern from Padding (PPP) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and tests in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes.

We present InfinityGAN, a method to generate arbitrary-sized images. The problem is associated with several key challenges. First, scaling existing models to an arbitrarily large image size is resource-constrained, both in terms of computation and availability of large-field-of-view training data. InfinityGAN trains and infers patch-by-patch seamlessly with low computational resources. Second, large images should be locally and globally consistent, avoid repetitive patterns, and look realistic. To address these, InfinityGAN takes global appearance, local structure and texture into account. With this formulation, we can generate images with spatial size and level of detail not attainable before. Experimental evaluation supports that InfinityGAN generates images with superior global structure compared to baselines and features parallelizable inference. Finally, we show several applications unlocked by our approach, such as fusing styles spatially, multi-modal outpainting and image inbetweening at arbitrary input and output sizes.

Image outpainting seeks for a semantically consistent extension of the input image beyond its available content. Compared to inpainting -- filling in missing pixels in a way coherent with the neighboring pixels -- outpainting can be achieved in more diverse ways since the problem is less constrained by the surrounding pixels. Existing image outpainting methods pose the problem as a conditional image-to-image translation task, often generating repetitive structures and textures by replicating the content available in the input image. In this work, we formulate the problem from the perspective of inverting generative adversarial networks. Our generator renders micro-patches conditioned on their joint latent code as well as their individual positions in the image. To outpaint an image, we seek for multiple latent codes not only recovering available patches but also synthesizing diverse outpainting by patch-based generation. This leads to richer structure and content in the outpainted regions. Furthermore, our formulation allows for outpainting conditioned on the categorical input, thereby enabling flexible user controls. Extensive experimental results demonstrate the proposed method performs favorably against existing in- and outpainting methods, featuring higher visual quality and diversity.

Conventional Neural Architecture Search (NAS) aims at finding a single architecture that achieves the best performance, which usually optimizes task related learning objectives such as accuracy. However, a single architecture may not be representative enough for the whole dataset with high diversity and variety. Intuitively, electing domain-expert architectures that are proficient in domain-specific features can further benefit architecture related objectives such as latency. In this paper, we propose InstaNAS---an instance-aware NAS framework---that employs a controller trained to search for a "distribution of architectures" instead of a single architecture; This allows the model to use sophisticated architectures for the difficult samples, which usually comes with large architecture related cost, and shallow architectures for those easy samples. During the inference phase, the controller assigns each of the unseen input samples with a domain expert architecture that can achieve high accuracy with customized inference costs. Experiments within a search space inspired by MobileNetV2 show InstaNAS can achieve up to 48.8% latency reduction without compromising accuracy on a series of datasets against MobileNetV2.

Humans can only interact with part of the surrounding environment due to biological restrictions. Therefore, we learn to reason the spatial relationships across a series of observations to piece together the surrounding environment. Inspired by such behavior and the fact that machines also have computational constraints, we propose COnditional COordinate GAN (COCO-GAN) of which the generator generates images by parts based on their spatial coordinates as the condition. On the other hand, the discriminator learns to justify realism across multiple assembled patches by global coherence, local appearance, and edge-crossing continuity. Despite the full images are never generated during training, we show that COCO-GAN can produce \textbf{state-of-the-art-quality} full images during inference. We further demonstrate a variety of novel applications enabled by teaching the network to be aware of coordinates. First, we perform extrapolation to the learned coordinate manifold and generate off-the-boundary patches. Combining with the originally generated full image, COCO-GAN can produce images that are larger than training samples, which we called "beyond-boundary generation". We then showcase panorama generation within a cylindrical coordinate system that inherently preserves horizontally cyclic topology. On the computation side, COCO-GAN has a built-in divide-and-conquer paradigm that reduces memory requisition during training and inference, provides high-parallelism, and can generate parts of images on-demand.

While image manipulation achieves tremendous breakthroughs (e.g., generating realistic faces) in recent years, video generation is much less explored and harder to control, which limits its applications in the real world. For instance, video editing requires temporal coherence across multiple clips and thus poses both start and end constraints within a video sequence. We introduce point-to-point video generation that controls the generation process with two control points: the targeted start- and end-frames. The task is challenging since the model not only generates a smooth transition of frames, but also plans ahead to ensure that the generated end-frame conforms to the targeted end-frame for videos of various length. We propose to maximize the modified variational lower bound of conditional data likelihood under a skip-frame training strategy. Our model can generate sequences such that their end-frame is consistent with the targeted end-frame without loss of quality and diversity. Extensive experiments are conducted on Stochastic Moving MNIST, Weizmann Human Action, and Human3.6M to evaluate the effectiveness of the proposed method. We demonstrate our method under a series of scenarios (e.g., dynamic length generation) and the qualitative results showcase the potential and merits of point-to-point generation.

The complementary characteristics of active and passive depth sensing techniques motivate the fusion of the Li-DAR sensor and stereo camera for improved depth perception. Instead of directly fusing estimated depths across LiDAR and stereo modalities, we take advantages of the stereo matching network with two enhanced techniques: Input Fusion and Conditional Cost Volume Normalization (CCVNorm) on the LiDAR information. The proposed framework is generic and closely integrated with the cost volume component that is commonly utilized in stereo matching neural networks. We experimentally verify the efficacy and robustness of our method on the KITTI Stereo and Depth Completion datasets, obtaining favorable performance against various fusion strategies. Moreover, we demonstrate that, with a hierarchical extension of CCVNorm, the proposed method brings only slight overhead to the stereo matching network in terms of computation time and model size.

Generative adversarial networks (GANs) often suffer from unpredictable mode-collapsing during training. We study the issue of mode collapse of Boundary Equilibrium Generative Adversarial Network (BEGAN), which is one of the state-of-the-art generative models. Despite its potential of generating high-quality images, we find that BEGAN tends to collapse at some modes after a period of training. We propose a new model, called BEGAN with a Constrained Space (BEGAN-CS), which includes a latent-space constraint in the loss function. We show that BEGAN-CS can significantly improve training stability and suppress mode collapse without either increasing the model complexity or degrading the image quality. Further, we visualize the distribution of latent vectors to elucidate the effect of latent-space constraint. The experimental results show that our method has additional advantages of being able to train on small datasets and to generate images similar to a given real image yet with variations of designated attributes on-the-fly.