A model learns object representations for physical scene understanding without direct supervision.
problem Learning object-centric representations without direct supervision of object properties.
method Object-Oriented Prediction and Planning (O2P2) model that learns perception, physics interaction, and rendering functions.
result The model can predict physical interactions and build block towers more complex than those seen during training.
Efficient model for foggy scene understanding in vehicles.
problem Challenging scene understanding and segmentation under foggy conditions.
method Domain adaptation and illumination-invariant image transformation.
result Outperforms state-of-the-art models in foggy scene understanding.
We propose a systematic learning-based approach to the generation of massive quantities of synthetic 3D scenes and arbitrary numbers of photorealistic 2D images thereof, with associated ground truth information, for the purposes of training, benchmarking, and diagnosing learning-based computer vision and robotics algor…
New text-to-image diffusion models improve scene understanding for AI agents.
problem Fine-grained scene understanding for AI agents from text and images.
method Pre-trained text-to-image diffusion models optimized for generating images from text prompts.
result Policies learned with Stable Control Representations outperform state-of-the-art approaches on various control tasks.
New model separates objects in scenes, enabling novel arrangements and depth.
problem Lack of modular, compositional scene modeling in generative models.
method Ensemble of generative models (experts) compete for explaining different parts of a scene.
result Model generates scenes with novel object arrangement and depth ordering.
Understanding how images of objects and scenes behave in response to specific ego-motions is a crucial aspect of proper visual development, yet existing visual learning methods are conspicuously disconnected from the physical source of their images. We propose to exploit proprioceptive motor signals to provide unsuperv…
New dataset tests mental rotation from single images, improving model understanding of 3D scenes.
problem Understanding how a scene looks from a different viewpoint using a single image.
method Created CLEVR-MRT dataset, explored neural architectures for volumetric scene representations.
result Demonstrated the effectiveness of volumetric representations in answering mental rotation questions.
Proposes Deep Scenes for interaction-aware scene understanding in reinforcement learning for autonomous driving.
problem Leveraging deep reinforcement learning for high-level decision making in autonomous driving requires handling variable-length sequences of different object types and interactions.
method Introduces Deep Scenes architecture, an extension of Deep Sets or Graph Convolutional Networks, to learn complex interaction-aware scene representations.
result Graph-Q and DeepScene-Q algorithms outperform state-of-the-art methods in evaluations with SUMO.
Deep neural networks decode natural visual scenes from neural spikes.
problem Decoding visual scenes from neural spikes for brain-machine interfaces.
method Developed a novel spike-image decoder (SID) using deep neural networks.
result SID reconstructs natural visual scenes from neural spikes with high accuracy.
Deep learning within the context of point clouds has gained much research interest in recent years mostly due to the promising results that have been achieved on a number of challenging benchmarks, such as 3D shape recognition and scene semantic segmentation. In many realistic settings however, snapshots of the environ…
Generative Multisensory Network learns 3D scene representations from multiple modalities.
problem Learning robust 3D scene representations from multiple sensory modalities.
method Amortized Product-of-Experts for efficient inference and cross-modal generation.
result The model can infer modality-invariant 3D scene representations efficiently from various sensory modalities.
Real-time scene understanding solved using Approximate Bayesian Computation.
problem Predicting human actions, object poses, and pedestrian crossings from depth images.
method Bayesian error model, neural surrogates, and adaptive discretization.
result Real-time inference on real-world problems is feasible.
Understanding and interacting with everyday physical scenes requires rich knowledge about the structure of the world, represented either implicitly in a value or policy function, or explicitly in a transition model. Here we introduce a new class of learnable models--based on graph networks--which implement an inductive…
Soft geometric bias improves physical dynamics predictions.
problem Learning physical dynamics with exact group equivariance can degrade performance.
method Object-centric world models using geometric algebra neural networks.
result Soft geometric inductive bias leads to better physical fidelity predictions.
The paper proposes a model to forecast traffic motion from sensor data.
problem Accurately predicting traffic motion for safe vehicle maneuvers.
method Implicit latent variable model using interaction graphs and graph neural networks.
result Achieves state-of-the-art motion forecasting and interaction understanding.
Scene understanding remains a significant challenge in the computer vision community. The visual psychophysics literature has demonstrated the importance of interdependence among parts of the scene. Yet, the majority of methods in computer vision remain local. Pictorial structures have arisen as a fundamental parts-bas…
Scene text magnifier enhances readability for visually impaired.
problem Helps visually impaired read natural scene text.
method Four CNN-based networks: character erasing, extraction, magnify, synthesis.
result Effective text magnification without background alteration.
A new method to decompose audio scenes using deep learning.
problem Understanding audio scenes with random microphone arrangements.
method Formulated a neural network for nonnegative tensor factorization.
result Learned sources' individual spectral dictionaries and activation patterns.
A novel method for visual question answering using scene graphs and reinforcement learning.
problem Answering free-form questions about images with deep linguistic and visual understanding.
method Context-driven, sequential reasoning based on scene graphs and reinforcement learning.
result Our method almost reaches human performance on the GQA dataset.
This work demonstrates a physical attack on a deep learning image classification system using projected light onto a physical scene. Prior work is dominated by techniques for creating adversarial examples which directly manipulate the digital input of the classifier. Such an attack is limited to scenarios where the adv…
Pix2Shape learns 3D scene representations from single images without supervision.
problem Learning 3D scene information from a single image without supervision.
method Pix2Shape uses an encoder, decoder, and critic network to generate 2.5D surfel-based reconstructions.
result Pix2Shape can generate complex 3D scenes from a single image, scaling with on-screen resolution.
Machine understanding of complex images is a key goal of artificial intelligence. One challenge underlying this task is that visual scenes contain multiple inter-related objects, and that global context plays an important role in interpreting the scene. A natural modeling framework for capturing such effects is structu…
Deep network predicts action sequences for complex tasks from a scene image.
problem Scalable task and motion planning from initial scene images.
method Deep convolutional recurrent neural network that predicts action sequences.
result Predicts promising action sequences, reducing motion planning problems.
Polarimetric images enhance object detection in adverse weather conditions.
problem Object detection in road scenes is challenging in adverse weather conditions.
method Combining polarimetric imaging and deep learning.
result Polarimetry improves object detection by 20% to 50% compared to conventional RGB images.
Real-life control tasks involve matters of various substances---rigid or soft bodies, liquid, gas---each with distinct physical behaviors. This poses challenges to traditional rigid-body physics engines. Particle-based simulators have been developed to model the dynamics of these complex scenes; however, relying on app…
BPI models 2D patterns on multiple planes and 3D scene from a single image.
problem Understanding and editing images with multiple 2D planes and 3D scene from a single image.
method Box Program Induction (BPI) with neural networks and search-based algorithm.
result Holistic, structured scene representation enables 3D-aware image editing.
Visual relationship detection can bridge the gap between computer vision and natural language for scene understanding of images. Different from pure object recognition tasks, the relation triplets of subject-predicate-object lie on an extreme diversity space, such as \textit{person-behind-person} and \textit{car-behind…
Generative model improves scene graph prediction for rare compositions.
problem Challenges in inferring unseen or rare scene graph compositions.
method Synthesizing rare scene graphs by perturbing real ones, using conditional GANs.
result Marginal but consistent improvements in zero- and few-shot metrics.
This work trains a model to predict human driving directions from road scenes.
problem Defining implicit rules of human behavior for autonomous vehicles.
method Self-supervised learning of probabilistic network model.
result Model successfully generalizes to new road scenes.
Study shows bottlenecks improve image segmentation quality.
problem Robust object discovery in real-world images remains challenging.
method Empirical investigation of reconstruction bottlenecks in GENESIS model.
result Reconstruction bottlenecks determine reconstruction and segmentation quality.
This work proposes a method to compose visual relations more faithfully.
problem Composing relations between objects in images is challenging due to their entanglement.
method Represent each relation as an unnormalized density (energy-based model) to compose relations factorizedly.
result The proposed method generates and edits scenes with multiple sets of relations more faithfully.
Physical patches can fool object detectors anywhere in an image.
problem Physical adversarial attacks on object detection systems.
method Demonstrated a patch that suppresses all detected objects regardless of its position.
result A patch can be placed anywhere in an image to suppress all detected objects, including those far away.
A central challenge in neuroscience is to understand neural computations and circuit mechanisms that underlie the encoding of ethologically relevant, natural stimuli. In multilayered neural circuits, nonlinear processes such as synaptic transmission and spiking dynamics present a significant obstacle to the creation of…
GIBLy adds geometric priors to 3D segmentation models, improving performance with minimal overhead.
problem Lack of explicit geometric information in 3D semantic segmentation models.
method Introduces GIBLy, a lightweight geometric inductive bias layer that integrates learnable geometric priors into existing 3D segmentation pipelines.
result Consistent performance gains across multiple benchmarks, including up to +11.5% mIoU on TS40K with PTV3.
NNs accurately predict energy eigenvalues and other physical phenomena in 1D quantum mechanics.
problem Understanding how neural networks interpret physics.
method Training NNs to predict energy eigenvalues from potentials and testing their ability to generalize.
result NNs can predict physical phenomena not learned during training, indicating a new way of understanding physics.
Analyzes GAN units for better understanding and improvement.
problem Lack of understanding GAN internal representations and artifacts.
method Interpretable units identification, causal effect quantification, contextual relationship examination.
result Visualizes and understands GANs at various levels, enabling new insights and improvements.
Neural networks are easier to code than understand, suggesting a focus on learning and development in neuroscience.
problem Difficulty in understanding neural network properties after training.
method Analogy with brain development and learning.
result Neural network properties are hard to understand compared to their training code.
Enhances neural rendering with geometry-aware attention.
problem Efficiently modeling complex 3D scenes.
method Introduces Epipolar Cross Attention (ECA) for non-local operations.
result Significant improvement in Generative Query Networks (GQN) performance.
A neural scene representation framework enforcing 3D transformations.
problem Learning 3D scene representations from images without 3D supervision.
method Introducing a loss enforcing equivariance of the scene representation with 3D transformations.
result Real-time neural rendering with comparable results to models requiring minutes for inference.
While current deep learning systems excel at tasks such as object classification, language processing, and gameplay, few can construct or modify a complex system such as a tower of blocks. We hypothesize that what these systems lack is a "relational inductive bias": a capacity for reasoning about inter-object relations…
BootsTAP uses real-world data to improve TAP tracking performance.
problem Lack of large-scale groundtruth training data for TAP.
method Bootstrapped training with a self-supervised student-teacher setup.
result State-of-the-art performance on TAP-Vid benchmarks.
DNCF framework recovers real scenes from imperfect images robustly.
problem Recovering real scenes from imperfect images.
method Nonparametric deep network that learns physical image formation equations.
result DNCF framework robustly defends against adversarial attacks.
OBSER framework infers sub-environments from objects, outperforming scene-based methods.
problem Zero-shot recognition of environments from object distributions.
method Bayesian framework using metric and self-supervised learning models to estimate object distributions in latent space.
result OBSER framework reliably performs inference in open-world and photorealistic environments, outperforming scene-based methods.
Maps are an important medium that enable people to comprehensively understand the configuration of cultural activities and natural elements over different times and places. Although massive maps are available in the digital era, how to effectively and accurately access the required map remains a challenge today. Previo…
GENESIS generates and samples 3D scenes by capturing object interactions.
problem Lack of models that explicitly capture object interactions in scene generation.
method Object-centric latent variables, spatial GMM, amortized inference, autoregressive prior.
result First object-centric generative model of 3D visual scenes.
We consider the problem of naming objects in complex, natural scenes containing widely varying object appearance and subtly different names. Informed by cognitive research, we propose an approach based on sharing context based object hypotheses between visual and lexical spaces. To this end, we present the Visual Seman…
Many machine learning image classifiers are vulnerable to adversarial attacks, inputs with perturbations designed to intentionally trigger misclassification. Current adversarial methods directly alter pixel colors and evaluate against pixel norm-balls: pixel perturbations smaller than a specified magnitude, according t…
ROOTS learns to represent and render 3D scenes with object-centric models.
problem Learning to represent and render 3D scenes with object-centric compositionality.
method Probabilistic generative model for learning object representations and scene rendering from partial observations.
result The model can infer 3D object representations and render scenes from arbitrary viewpoints.