In this paper we propose a neural message passing approach to augment an input 3D indoor scene with new objects matching their surroundings. Given an input, potentially incomplete, 3D scene and a query location, our method predicts a probability distribution over object types that fit well in that location. Our distribution is predicted though passing learned messages in a dense graph whose nodes represent objects in the input scene and edges represent spatial and structural relationships. By weighting messages through an attention mechanism, our method learns to focus on the most relevant surrounding scene context to predict new scene objects. We found that our method significantly outperforms state-of-the-art approaches in terms of correctly predicting objects missing in a scene based on our experiments in the SUNCG dataset. We also demonstrate other applications of our method, including context-based 3D object recognition and iterative scene generation.
Recently there is an increasing interest in scene generation within the research community. However, models used for generating scene layouts from textual description largely ignore plausible visual variations within the structure dictated by the text. We propose LayoutVAE, a variational autoencoder based framework for generating stochastic scene layouts. LayoutVAE is a versatile modeling framework that allows for generating full image layouts given a label set, or per label layouts for an existing image given a new label. In addition, it is also capable of detecting unusual layouts, potentially providing a way to evaluate layout generation problem. Extensive experiments on MNIST-Layouts and challenging COCO 2017 Panoptic dataset verifies the effectiveness of our proposed framework.
3D multi object generative models allow us to synthesize a large range of novel 3D multi object scenes and also identify objects, shapes, layouts and their positions. But multi object scenes are difficult to create because of the dataset being multimodal in nature. The conventional 3D generative adversarial models are not efficient in generating multi object scenes, they usually tend to generate either one object or generate fuzzy results of multiple objects. Auto-encoder models have much scope in feature extraction and representation learning using the unsupervised paradigm in probabilistic spaces. We try to make use of this property in our proposed model. In this paper we propose a novel architecture using 3DConvNets trained with the progressive training paradigm that has been able to generate realistic high resolution 3D scenes of rooms, bedrooms, offices etc. with various pieces of furniture and objects. We make use of the adversarial auto-encoder along with the WGAN-GP loss parameter in our discriminator loss function. Finally this new approach to multi object scene generation has also been able to generate more number of objects per scene.
The visual world we sense, interpret and interact everyday is a complex composition of interleaved physical entities. Therefore, it is a very challenging task to generate vivid scenes of similar complexity using computers. In this work, we present a scene generation framework based on Generative Adversarial Networks (GANs) to sequentially compose a scene, breaking down the underlying problem into smaller ones. Different than the existing approaches, our framework offers an explicit control over the elements of a scene through separate background and foreground generators. Starting with an initially generated background, foreground objects then populate the scene one-by-one in a sequential manner. Via quantitative and qualitative experiments on a subset of the MS-COCO dataset, we show that our proposed framework produces not only more diverse images but also copes better with affine transformations and occlusion artifacts of foreground objects than its counterparts.
Layout is important for graphic design and scene generation. We propose a novel Generative Adversarial Network, called LayoutGAN, that synthesizes layouts by modeling geometric relations of different types of 2D elements. The generator of LayoutGAN takes as input a set of randomly-placed 2D graphic elements and uses self-attention modules to refine their labels and geometric parameters jointly to produce a realistic layout. Accurate alignment is critical for good layouts. We thus propose a novel differentiable wireframe rendering layer that maps the generated layout to a wireframe image, upon which a CNN-based discriminator is used to optimize the layouts in image space. We validate the effectiveness of LayoutGAN in various experiments including MNIST digit generation, document layout generation, clipart abstract scene generation and tangram graphic design.
We propose a new probabilistic programming language for the design and analysis of perception systems, especially those based on machine learning. Specifically, we consider the problems of training a perception system to handle rare events, testing its performance under different conditions, and debugging failures. We show how a probabilistic programming language can help address these problems by specifying distributions encoding interesting types of inputs and sampling these to generate specialized training and test sets. More generally, such languages can be used for cyber-physical systems and robotics to write environment models, an essential prerequisite to any formal analysis. In this paper, we focus on systems like autonomous cars and robots, whose environment is a "scene", a configuration of physical objects and agents. We design a domain-specific language, Scenic, for describing "scenarios" that are distributions over scenes. As a probabilistic programming language, Scenic allows assigning distributions to features of the scene, as well as declaratively imposing hard and soft constraints over the scene. We develop specialized techniques for sampling from the resulting distribution, taking advantage of the structure provided by Scenic's domain-specific syntax. Finally, we apply Scenic in a case study on a convolutional neural network designed to detect cars in road images, improving its performance beyond that achieved by state-of-the-art synthetic data generation methods.
The ability to map descriptions of scenes to 3D geometric representations has many applications in areas such as art, education, and robotics. However, prior work on the text to 3D scene generation task has used manually specified object categories and language that identifies them. We introduce a dataset of 3D scenes annotated with natural language descriptions and learn from this data how to ground textual descriptions to physical objects. Our method successfully grounds a variety of lexical terms to concrete referents, and we show quantitatively that our method improves 3D scene generation over previous work using purely rule-based methods. We evaluate the fidelity and plausibility of 3D scenes generated with our grounding approach through human judgments. To ease evaluation on this task, we also introduce an automated metric that strongly correlates with human judgments.
In many robotic domains such as flexible automated manufacturing or personal assistance, a fundamental perception task is that of identifying and localizing objects whose 3D models are known. Canonical approaches to this problem include discriminative methods that find correspondences between feature descriptors computed over the model and observed data. While these methods have been employed successfully, they can be unreliable when the feature descriptors fail to capture variations in observed data; a classic cause being occlusion. As a step towards deliberative reasoning, we present PERCH: PErception via SeaRCH, an algorithm that seeks to find the best explanation of the observed sensor data by hypothesizing possible scenes in a generative fashion. Our contributions are: i) formulating the multi-object recognition and localization task as an optimization problem over the space of hypothesized scenes, ii) exploiting structure in the optimization to cast it as a combinatorial search problem on what we call the Monotone Scene Generation Tree, and iii) leveraging parallelization and recent advances in multi-heuristic search in making combinatorial search tractable. We prove that our system can guaranteedly produce the best explanation of the scene under the chosen cost function, and validate our claims on real world RGB-D test data. Our experimental results show that we can identify and localize objects under heavy occlusion--cases where state-of-the-art methods struggle.
Designing 3D scenes is currently a creative task that requires significant expertise and effort in using complex 3D design interfaces. This effortful design process starts in stark contrast to the easiness with which people can use language to describe real and imaginary environments. We present SceneSeer: an interactive text to 3D scene generation system that allows a user to design 3D scenes using natural language. A user provides input text from which we extract explicit constraints on the objects that should appear in the scene. Given these explicit constraints, the system then uses a spatial knowledge base learned from an existing database of 3D scenes and 3D object models to infer an arrangement of the objects forming a natural scene matching the input description. Using textual commands the user can then iteratively refine the created scene by adding, removing, replacing, and manipulating objects. We evaluate the quality of 3D scenes generated by SceneSeer in a perceptual evaluation experiment where we compare against manually designed scenes and simpler baselines for 3D scene generation. We demonstrate how the generated scenes can be iteratively refined through simple natural language commands.
Generalization performance of trained computer vision systems that use computer graphics (CG) generated data is not yet effective due to the concept of 'domain-shift' between virtual and real data. Although simulated data augmented with a few real world samples has been shown to mitigate domain shift and improve transferability of trained models, guiding or bootstrapping the virtual data generation with the distributions learnt from target real world domain is desired, especially in the fields where annotating even few real images is laborious (such as semantic labeling, and intrinsic images etc.). In order to address this problem in an unsupervised manner, our work combines recent advances in CG (which aims to generate stochastic scene layouts coupled with large collections of 3D object models) and generative adversarial training (which aims train generative models by measuring discrepancy between generated and real data in terms of their separability in the space of a deep discriminatively-trained classifier). Our method uses iterative estimation of the posterior density of prior distributions for a generative graphical model. This is done within a rejection sampling framework. Initially, we assume uniform distributions as priors on the parameters of a scene described by a generative graphical model. As iterations proceed the prior distributions get updated to distributions that are closer to the (unknown) distributions of target data. We demonstrate the utility of adversarially tuned scene generation on two real-world benchmark datasets (CityScapes and CamVid) for traffic scene semantic labeling with a deep convolutional net (DeepLab). We realized performance improvements by 2.28 and 3.14 points (using the IoU metric) between the DeepLab models trained on simulated sets prepared from the scene generation models before and after tuning to CityScapes and CamVid respectively.
Key to automatically generate natural scene images is to properly arrange among various spatial elements, especially in the depth direction. To this end, we introduce a novel depth structure preserving scene image generation network (DSP-GAN), which favors a hierarchical and heterogeneous architecture, for the purpose of depth structure preserving scene generation. The main trunk of the proposed infrastructure is built on a Hawkes point process that models the spatial dependency between different depth layers. Within each layer generative adversarial sub-networks are trained collaboratively to generate realistic scene components, conditioned on the layer information produced by the point process. We experiment our model on a sub-set of SUNdataset with annotated scene images and demonstrate that our models are capable of generating depth-realistic natural scene image.
Learning to generate natural scenes has always been a challenging task in computer vision. It is even more painstaking when the generation is conditioned on images with drastically different views. This is mainly because understanding, corresponding, and transforming appearance and semantic information across the views is not trivial. In this paper, we attempt to solve the novel problem of cross-view image synthesis, aerial to street-view and vice versa, using conditional generative adversarial networks (cGAN). Two new architectures called Crossview Fork (X-Fork) and Crossview Sequential (X-Seq) are proposed to generate scenes with resolutions of 64x64 and 256x256 pixels. X-Fork architecture has a single discriminator and a single generator. The generator hallucinates both the image and its semantic segmentation in the target view. X-Seq architecture utilizes two cGANs. The first one generates the target image which is subsequently fed to the second cGAN for generating its corresponding semantic segmentation map. The feedback from the second cGAN helps the first cGAN generate sharper images. Both of our proposed architectures learn to generate natural images as well as their semantic segmentation maps. The proposed methods show that they are able to capture and maintain the true semantics of objects in source and target views better than the traditional image-to-image translation method which considers only the visual appearance of the scene. Extensive qualitative and quantitative evaluations support the effectiveness of our frameworks, compared to two state of the art methods, for natural scene generation across drastically different views.
Convolutional Neural Networks (CNNs) have begun to permeate all corners of electronic society (from voice recognition to scene generation) due to their high accuracy and machine efficiency per operation. At their core, CNN computations are made up of multi-dimensional dot products between weight and input vectors. This paper studies how weight repetition ---when the same weight occurs multiple times in or across weight vectors--- can be exploited to save energy and improve performance during CNN inference. This generalizes a popular line of work to improve efficiency from CNN weight sparsity, as reducing computation due to repeated zero weights is a special case of reducing computation due to repeated weights. To exploit weight repetition, this paper proposes a new CNN accelerator called the Unique Weight CNN Accelerator (UCNN). UCNN uses weight repetition to reuse CNN sub-computations (e.g., dot products) and to reduce CNN model size when stored in off-chip DRAM ---both of which save energy. UCNN further improves performance by exploiting sparsity in weights. We evaluate UCNN with an accelerator-level cycle and energy model and with an RTL implementation of the UCNN processing element. On three contemporary CNNs, UCNN improves throughput-normalized energy consumption by 1.2x - 4x, relative to a similarly provisioned baseline accelerator that uses Eyeriss-style sparsity optimizations. At the same time, the UCNN processing element adds only 17-24% area overhead relative to the same baseline.
본 논문에서는 기반 생성 파이프라인을 수정하지 않고도 부분적인 기하학적 관측이 주어졌을 때 생성된 3D 물체를 장면에 배치하는, 최신 네이티브 형상 생성 모델로 생성된 물체의 씬 레벨(scene-level) 포즈 추정 프레임워크인 ScenePlacer를 제시한다. 기존의 장면 생성 방식들은 정규 공간(canonical space) 사전 확률(prior)을 저해하며 물체를 장면 좌표계에서 직접 생성하거나, 형상 생성 과정에서 완전한 생성 형상으로부터 포즈를 회귀한다. 후자의 경우 스케일 모호성과 부분적 관측에 민감한 불량 조건(ill-conditioned) 문제이다. ScenePlacer는 정렬을 조건부 가시성 필드(visibility field) 학습으로 재구성함으로써 물체 생성과 포즈 추정을 분리한다. 복원 결과로부터 얻은 관측된 부분 포인트 클라우드와 생성된 메시가 주어지면, 물체 공간에서 정의된 가시성 필드를 예측한 다음, 부분 관측을 메시의 가시 영역과 정렬하여 강체 변환(rigid transformation)을 추정한다. 본 연구에서는 사전학습된 3D 형상 VAE를 백본으로 제안하여, 별도로 인코딩된 관측 정보에 조건을 부여하면서도 메시 사전 확률을 보존한다. 나아가 가시성을 인식하는 반경 제한(radius-constrained) 지도 방식을 도입하고, 물체 수준의 메시와 렌더링된 뷰만을 사용하여 학습 쌍을 합성함으로써 씬 레벨 포즈 어노테이션의 필요성을 제거한다. 3D-FUTURE 데이터셋에서 ScenePlacer는 참값(ground-truth) 및 단안 깊이 추정 환경 모두에서 기존의 배치 및 포즈 정제(pose-refinement) 베이스라인들을 일관되게 능가한다. 생성된 물체를 사용할 때 추정된 깊이 환경에서 0.0077의 Chamfer 거리와 92.40의 F-score를 달성하며, 참값 형상 환경에서는 각각 0.0023 및 98.27로 향상된다.
3D 장면 생성은 몰입형 미디어, 로보틱스, 자율주행, 체화 AI(embodied AI) 등의 응용을 위해 공간적으로 구조화되고 의미론적으로 유의미하며 포토리얼리스틱(photorealistic)한 환경을 합성하는 것을 목표로 한다. 절차적 규칙(procedural rules)에 기반한 초기 방법은 확장성은 제공했으나 다양성은 제한적이었다. 딥 생성 모델(예: GAN, diffusion model)과 3D 표현(예: NeRF, 3D Gaussians)의 최근 발전은 실세계 장면 분포의 학습을 가능하게 하여 충실도, 다양성, 시점 일관성을 개선했다. diffusion model과 같은 최근 기법은 생성을 이미지 또는 영상 합성 문제로 재구성함으로써 3D 장면 합성과 포토리얼리즘(photorealism)을 연결한다. 본 조사는 최신(state-of-the-art) 접근법에 대한 체계적 개관을 제공하며, 이를 절차적 생성, 신경망 3D 기반 생성, 이미지 기반 생성, 영상 기반 생성의 네 가지 패러다임으로 구분한다. 각 방법의 기술적 기반, 트레이드오프, 대표적 결과를 분석하고, 널리 사용되는 데이터셋, 평가(evaluation) 프로토콜, 하류 응용을 검토한다. 마지막으로 생성 역량, 3D 표현, 데이터 및 주석, 평가의 핵심 과제를 논의하고, 더 높은 충실도, 물리 인식(physics-aware) 및 대화형(interactive) 생성, 통합 지각-생성 모델을 포함한 유망한 연구 방향을 개괄한다. 본 리뷰는 3D 장면 생성의 최근 진전을 정리하고 생성 AI, 3D 비전, 체화 지능의 교차점에서의 유망한 방향을 강조한다. 진행 중인 동향을 추적하기 위해 최신 프로젝트 페이지를 유지한다: https://github.com/hzxie/Awesome-3D-Scene-Generation.
TabletopGen은 텍스트 또는 단일 이미지 입력만으로 인스턴스 수준의 상호작용과 물리적 타당성을 갖춘 탁상(tabletop) 3D 장면을 자동 생성하는 학습 불필요(training-free) 시뮬레이션 엔진으로, Differentiable Rotation Optimizer와 Top-View Spatial Alignment를 결합한 새로운 자세·스케일 정렬 방법을 통해 충돌 없는 레이아웃을 복원한다. 생성된 장면은 물리 시뮬레이터에 직접 조립되어 로봇 조작 데이터 합성에 활용되며, 제로샷 real-to-sim-to-real 정책 전이 실험을 통해 실제 로봇 팔에서의 실행 가능성이 검증되었다.
Generating explorable 3D scenes from a single image requires strong generative priors and accurate geometric representations suitable for downstream use. Current video diffusion models offer high-quality generation and implicitly encode multi-view geometric structure in latent space. However, existing feedforward latent scene decoders typically output volumetric 3D Gaussians that lack a well-defined surface, limiting their use in simulation or standard graphics pipelines. This motivates decoding surface-aligned primitives that are not only renderable but also closer to explicit geometric assets. We ask whether compressed video diffusion latents can be mapped directly to explicit surface primitives in a single pass. To this end, we introduce FLAT and, for the first time, show that triangle splats can be decoded directly from video diffusion latents. Compared with decoding 3D Gaussians, predicting flat primitives is notoriously more challenging due to high sensitivity to primitive orientations, oftentimes leading to poor gradient flow. FLAT solves with two key ingredients: a ray-centered rotation parameterization for triangle regression and a novel product window function that improves gradient flow during differentiable triangle rendering. On standard benchmarks, FLAT achieves significantly better geometric accuracy while maintaining competitive visual quality compared to state-of-the-art feedforward baselines. We further show that a lightweight test-time refinement step converts the predicted triangle soup into a fully opaque, game-engine-ready representation that supports real-time rendering. By evaluating 3DGS, 2DGS, and triangle splatting variants under an identical training setup, we provide the first systematic analysis of representation tradeoffs in feedforward scene generation. The project page is available at https://flat-splat.github.io