We present Text2Room, a method for generating room-scale textured 3D meshes from a given text prompt as input. To this end, we leverage pre-trained 2D text-to-image models to synthesize a sequence of images from different poses. In order to lift these outputs into a consistent 3D scene representation, we combine monocular depth estimation with a text-conditioned inpainting model. The core idea of our approach is a tailored viewpoint selection such that the content of each image can be fused into a seamless, textured 3D mesh. More specifically, we propose a continuous alignment strategy that iteratively fuses scene frames with the existing geometry to create a seamless mesh. Unlike existing works that focus on generating single objects or zoom-out trajectories from text, our method generates complete 3D scenes with multiple objects and explicit 3D geometry. We evaluate our approach using qualitative and quantitative metrics, demonstrating it as the first method to generate room-scale 3D geometry with compelling textures from only text as input.
Although an object may appear in numerous contexts, we often describe it in a limited number of ways. This happens because language abstracts away visual variation to represent and communicate concepts. Building on this intuition, we propose an alternative approach to visual learning: using language similarity to sample semantically similar image pairs for contrastive learning. Our approach deviates from image-based contrastive learning by using language to sample pairs instead of hand-crafted augmentations or learned clusters. Our approach also deviates from image-text contrastive learning by relying on pre-trained language models to guide the learning rather than minimize a cross-modal similarity. Through a series of experiments, we show that language-guided learning can learn better features than both image-image and image-text representation learning approaches.
We present MAV3D (Make-A-Video3D), a method for generating three-dimensional dynamic scenes from text descriptions. Our approach uses a 4D dynamic Neural Radiance Field (NeRF), which is optimized for scene appearance, density, and motion consistency by querying a Text-to-Video (T2V) diffusion-based model. The dynamic video output generated from the provided text can be viewed from any camera location and angle, and can be composited into any 3D environment. MAV3D does not require any 3D or 4D data and the T2V model is trained only on Text-Image pairs and unlabeled videos. We demonstrate the effectiveness of our approach using comprehensive quantitative and qualitative experiments and show an improvement over previously established internal baselines. To the best of our knowledge, our method is the first to generate 3D dynamic scenes given a text description.
Modeling and re-rendering dynamic 3D scenes is a challenging task in 3D vision. Prior approaches build on NeRF and rely on implicit representations. This is slow since it requires many MLP evaluations, constraining real-world applications. We show that dynamic 3D scenes can be explicitly represented by six planes of learned features, leading to an elegant solution we call HexPlane. A HexPlane computes features for points in spacetime by fusing vectors extracted from each plane, which is highly efficient. Pairing a HexPlane with a tiny MLP to regress output colors and training via volume rendering gives impressive results for novel view synthesis on dynamic scenes, matching the image quality of prior work but reducing training time by more than $100\times$. Extensive ablations confirm our HexPlane design and show that it is robust to different feature fusion mechanisms, coordinate systems, and decoding mechanisms. HexPlanes are a simple and effective solution for representing 4D volumes, and we hope they can broadly contribute to modeling spacetime for dynamic 3D scenes.
A central goal of visual recognition is to understand objects and scenes from a single image. 2D recognition has witnessed tremendous progress thanks to large-scale learning and general-purpose representations. Comparatively, 3D poses new challenges stemming from occlusions not depicted in the image. Prior works try to overcome these by inferring from multiple views or rely on scarce CAD models and category-specific priors which hinder scaling to novel settings. In this work, we explore single-view 3D reconstruction by learning generalizable representations inspired by advances in self-supervised learning. We introduce a simple framework that operates on 3D points of single objects or whole scenes coupled with category-agnostic large-scale training from diverse RGB-D videos. Our model, Multiview Compressive Coding (MCC), learns to compress the input appearance and geometry to predict the 3D structure by querying a 3D-aware decoder. MCC's generality and efficiency allow it to learn from large-scale and diverse data sources with strong generalization to novel objects imagined by DALL$\cdot$E 2 or captured in-the-wild with an iPhone.
3D shapes have complementary abstractions from low-level geometry to part-based hierarchies to languages, which convey different levels of information. This paper presents a unified framework to translate between pairs of shape abstractions: $\textit{Text}$ $\Longleftrightarrow$ $\textit{Point Cloud}$ $\Longleftrightarrow$ $\textit{Program}$. We propose $\textbf{Neural Shape Compiler}$ to model the abstraction transformation as a conditional generation process. It converts 3D shapes of three abstract types into unified discrete shape code, transforms each shape code into code of other abstract types through the proposed $\textit{ShapeCode Transformer}$, and decodes them to output the target shape abstraction. Point Cloud code is obtained in a class-agnostic way by the proposed $\textit{Point}$VQVAE. On Text2Shape, ShapeGlot, ABO, Genre, and Program Synthetic datasets, Neural Shape Compiler shows strengths in $\textit{Text}$ $\Longrightarrow$ $\textit{Point Cloud}$, $\textit{Point Cloud}$ $\Longrightarrow$ $\textit{Text}$, $\textit{Point Cloud}$ $\Longrightarrow$ $\textit{Program}$, and Point Cloud Completion tasks. Additionally, Neural Shape Compiler benefits from jointly training on all heterogeneous data and tasks.
Video provides us with the spatio-temporal consistency needed for visual learning. Recent approaches have utilized this signal to learn correspondence estimation from close-by frame pairs. However, by only relying on close-by frame pairs, those approaches miss out on the richer long-range consistency between distant overlapping frames. To address this, we propose a self-supervised approach for correspondence estimation that learns from multiview consistency in short RGB-D video sequences. Our approach combines pairwise correspondence estimation and registration with a novel SE(3) transformation synchronization algorithm. Our key insight is that self-supervised multiview registration allows us to obtain correspondences over longer time frames; increasing both the diversity and difficulty of sampled pairs. We evaluate our approach on indoor scenes for correspondence estimation and RGB-D pointcloud registration and find that we perform on-par with supervised approaches.
Most neural networks for computer vision are designed to infer using RGB images. However, these RGB images are commonly encoded in JPEG before saving to disk; decoding them imposes an unavoidable overhead for RGB networks. Instead, our work focuses on training Vision Transformers (ViT) directly from the encoded features of JPEG. This way, we can avoid most of the decoding overhead, accelerating data load. Existing works have studied this aspect but they focus on CNNs. Due to how these encoded features are structured, CNNs require heavy modification to their architecture to accept such data. Here, we show that this is not the case for ViTs. In addition, we tackle data augmentation directly on these encoded features, which to our knowledge, has not been explored in-depth for training in this setting. With these two improvements -- ViT and data augmentation -- we show that our ViT-Ti model achieves up to 39.2% faster training and 17.9% faster inference with no accuracy loss compared to the RGB counterpart.
We present a simple baseline for directly estimating the relative pose (rotation and translation, including scale) between two images. Deep methods have recently shown strong progress but often require complex or multi-stage architectures. We show that a handful of modifications can be applied to a Vision Transformer (ViT) to bring its computations close to the Eight-Point Algorithm. This inductive bias enables a simple method to be competitive in multiple settings, often substantially improving over the state of the art with strong performance gains in limited data regimes.
Recognizing scenes and objects in 3D from a single image is a longstanding goal of computer vision with applications in robotics and AR/VR. For 2D recognition, large datasets and scalable solutions have led to unprecedented advances. In 3D, existing benchmarks are small in size and approaches specialize in few object categories and specific domains, e.g. urban driving scenes. Motivated by the success of 2D recognition, we revisit the task of 3D object detection by introducing a large benchmark, called Omni3D. Omni3D re-purposes and combines existing datasets resulting in 234k images annotated with more than 3 million instances and 97 categories.3D detection at such scale is challenging due to variations in camera intrinsics and the rich diversity of scene and object types. We propose a model, called Cube R-CNN, designed to generalize across camera and scene types with a unified approach. We show that Cube R-CNN outperforms prior works on the larger Omni3D and existing benchmarks. Finally, we prove that Omni3D is a powerful dataset for 3D object recognition, show that it improves single-dataset performance and can accelerate learning on new smaller datasets via pre-training.