Planning allows an agent to safely refine its actions before executing them in the real world. In autonomous driving, this is crucial to avoid collisions and navigate in complex, dense traffic scenarios. One way to plan is to search for the best action sequence. However, this is challenging when all necessary components – policy, next-state predictor, and critic – have to be learned. Here we propose Differentiable Simulation for Search (DSS), a framework that leverages the differentiable simulator Waymax as both a next state predictor and a critic. It relies on the simulator’s hardcoded dynamics, making state predictions highly accurate, while utilizing the simulator’s differentiability to effectively search across action sequences. Our DSS agent optimizes its actions using gradient descent over imagined future trajectories. We show experimentally that DSS – the combination of planning gradients and stochastic search – significantly improves tracking and path planning accuracy compared to sequence prediction, imitation learning, model-free RL, and other planning methods.

Human activities are inherently complex, and even simple household tasks involve numerous object interactions. To better understand these activities, it is crucial to model their interactions with the environment captured through dynamic changes. The recent availability of affordable head-mounted cameras and egocentric data offers a more accessible and efficient means to understand dynamic human-object interactions in 3D environments. However, most existing methods for human activity modeling either focus on reconstructing 3D models of hand-object or human-scene interactions or on mapping 3D scenes, neglecting dynamic interactions with objects. The few existing solutions often require inputs from multiple sources, including multi-camera setups, depth-sensing cameras, or kinesthetic sensors. To this end, we introduce EgoGaussian, the first method capable of simultaneously reconstructing 3D scenes and dynamically tracking 3D object motion from RGB egocentric input alone. We leverage the uniquely discrete nature of Gaussian Splatting and segment dynamic interactions from the background. Our approach employs a clip-level online learning pipeline that leverages the dynamic nature of human activities, allowing us to reconstruct the temporal evolution of the scene in chronological order and track rigid object motion. Additionally, our method automatically segments object and background Gaussians, providing explicit 3D representations for both static scenes and dynamic objects. EgoGaussian shows significant improvements in terms of both dynamic object and background reconstruction quality compared to the state-of-the-art. We also qualitatively demonstrate the high quality of the reconstructed models.

We present GEM, a Generalizable Ego-vision Multimodal world model that predicts future frames using a reference frame, sparse features, human poses, and ego-trajectories. Hence, our model has precise control over object dynamics, ego-agent motion and human poses. GEM generates paired RGB and depth outputs for richer spatial understanding. We introduce autoregressive noise schedules to enable stable long-horizon generations. Our dataset is comprised of 4000+ hours of multimodal data across domains like autonomous driving, egocentric human activities, and drone flights. Pseudo-labels are used to get depth maps, ego-trajectories, and human poses. We use a comprehensive evaluation framework, including a new Control of Object Manipulation (COM) metric, to assess controllability. Experiments show GEM excels at generating diverse, controllable scenarios and temporal consistency over long generations. Code, models, and datasets are fully open-sourced1.

Modern world models require costly and time-consuming collection of large video datasets with action demonstrations by people or by environment-specific agents. To simplify training, we focus on using many virtual environments for inexpensive, automatically collected interaction data. Genie [5], a recent multi-environment world model, demonstrates simulation abilities of many environments with shared behavior. Unfortunately, training their model requires expensive demonstrations. Therefore, we propose a training framework merely using a random agent in virtual environments. While the model trained in this manner exhibits good controls, it is limited by the random exploration possibilities. To address this limitation, we propose AutoExplore Agent - an exploration agent that entirely relies on the uncertainty of the world model, delivering diverse data from which it can learn the best. Our agent is fully independent of environment-specific rewards and thus adapts easily to new environments. With this approach, the pretrained multi-environment model can quickly adapt to new environments achieving video fidelity and controllability improvement.In order to obtain automatically large-scale interaction datasets for pretraining, we group environments with similar behavior and controls. To this end, we annotate the behavior and controls of 974 virtual environments - a dataset that we name RetroAct. For building our model, we first create an open implementation of Genie - GenieRedux and apply enhancements and adaptations in our version GenieRedux-G. Our code and data are available at https://github.com/insait-institute/GenieRedux.

Understanding drivers’ decision-making is crucial for road safety. Although predicting the ego-vehicle’s path is valuable for driver-assistance systems, existing methods mainly focus on external factors like other vehicles’ motions, often neglecting the driver’s attention and intent. To address this gap, we infer the ego-trajectory by integrating the driver’s gaze and the surrounding scene. We introduce RouteFormer, a novel multimodal ego-trajectory prediction network combining GPS data, environmental context, and the driver's field-of-view—comprising first-person video and gaze fixations. We also present the Path Complexity Index (PCI), a new metric for trajectory complexity that enables a more nuanced evaluation of challenging scenarios. To tackle data scarcity and enhance diversity, we introduce GEM, a comprehensive dataset of urban driving scenarios enriched with synchronized driver field-of-view and gaze data. Extensive evaluations on GEM and DR(eye)VE demonstrate that RouteFormer significantly outperforms state-of-the-art methods, achieving notable improvements in prediction accuracy across diverse conditions. Ablation studies reveal that incorporating driver field-of-view data yields significantly better average displacement error, especially in challenging scenarios with high PCI scores, underscoring the importance of modeling driver attention. All data and code are available at meakbiyik.github.io/routeformer.

We propose LangHOPS, the first Multimodal Large Language Model (MLLM)-based framework for open-vocabulary object–part instance segmentation. Given an image, LangHOPS can jointly detect and segment hierarchical object and part instances from open-vocabulary candidate categories. Unlike prior approaches that rely on heuristic or learnable visual grouping, our approach grounds object–part hierarchies in language space. It integrates the MLLM into the object-part parsing pipeline to leverage rich knowledge and reasoning capabilities, and link multi-granularity concepts within the hierarchies. We evaluate LangHOPS across multiple challenging scenarios, including in-domain and cross-dataset object-part instance segmentation, and zero-shot semantic segmentation. LangHOPS achieves state-of-the-art results, surpassing previous methods by 5.5% Average Precision(AP) (in-domain) and 4.8% (cross-dataset) on the PartImageNet dataset and by 2.5% mIOU on unseen object parts in ADE20K (zero-shot). Ablation studies further validate the effectiveness of the language-grounded hierarchy and MLLM-driven part query refinement strategy.

World models have recently gained prominence for action-conditioned visual prediction in complex environments. However, relying on only a few recent observations causes them to lose long-term context. Consequently, within a few steps, the generated scenes drift from what was previously observed, undermining temporal coherence. This limitation, common in state-of-the-art world models, which are diffusion-based, stems from the lack of a lasting environment state. To address this problem, we introduce StateSpaceDiffuser, where a diffusion model is enabled to perform long-context tasks by integrating features from a state-space model, representing the entire interaction history. This design restores long-term memory while preserving the high-fidelity synthesis of diffusion models. To rigorously measure temporal consistency, we develop an evaluation protocol that probes a model’s ability to reinstantiate seen content in extended rollouts. Comprehensive experiments show that StateSpaceDiffuser significantly outperforms a strong diffusion-only baseline, maintaining a coherent visual context for an order of magnitude more steps. It delivers consistent views in both a 2D maze navigation and a complex 3D environment. These results establish that bringing state-space representations into diffusion models is highly effective in demonstrating both visual details and long-term memory. Project page: https://insait-institute.github.io/StateSpaceDiffuser/

Driver's eye gaze holds a wealth of cognitive and intentional cues crucial for intelligent vehicles. Despite its sig-nificance, research on in-vehicle gaze estimation remains limited due to the scarcity of comprehensive and well-annotated datasets in real driving scenarios. In this pa-per, we present three novel elements to advance in-vehicle gaze research. Firstly, we introduce IVGaze, a pioneering dataset capturing in-vehicle gaze, collected from 125 sub-jects and covering a large range of gaze and head poses within vehicles. In this dataset, we propose a new vision-based solution for in-vehicle gaze collection, introducing a refined gaze target calibration method to tackle annotation challenges. Second, our research focuses on in-vehicle gaze estimation leveraging the IvGaze. In-vehicle face images often suffer from low resolution, prompting our in-troduction of a gaze pyramid transformer that leverages transformer-based multilevel features integration. Expanding upon this, we introduce the dual-stream gaze pyramid transformer (GazeDPTR). Employing perspective transfor-mation, we rotate virtual cameras to normalize images, uti-lizing camera pose to merge normalized and original images for accurate gaze estimation. GazeDPTR shows state-of-the-art performance on the IVGaze dataset. Thirdly, we explore a novel strategy for gaze zone classification by extending the GazeDPTR. A foundational tri-plane and project gaze onto these planes are newly defined. Leveraging both positional features from the projection points and visual attributes from images, we achieve superior performance compared to relying solely on visual features, sub-stantiating the advantage of gaze estimation. The project is available at https://yihua.zone/work/ivgaze.

Synthesizing natural human motions that enable a 3D human avatar to walk and reach for arbitrary goals in 3D space remains an unsolved problem with many applications. Existing methods (data-driven or using reinforcement learning) are limited in terms of generalization and motion naturalness. A primary obstacle is the scarcity of training data that combines locomotion with goal reaching. To ad-dress this, we introduce WANDR, a data-driven model that takes an avatar's initial pose and a goal's 3D position and generates natural human motions that place the end effec-tor (wrist) on the goal location. To solve this, we intro-duce novel intention features that drive rich goal-oriented movement. Intention guides the agent to the goal, and in-teractively adapts the generation to novel situations without needing to define sub-goals or the entire motion path. Cru-cially, intention allows training on datasets that have goal-oriented motions as well as those that do not. WANDR is a conditional Variational Auto-Encoder (c- VAE), which we train using the AMASS and CIRCLE datasets. We evaluate our method extensively and demonstrate its ability to gener-ate natural and long-term motions that reach 3D goals and generalize to unseen goal locations. Our models and code are available for research purposes at wandr.is.tue.mpg.de.

We study object interaction anticipation in egocentric videos. This task requires an understanding of the spatio-temporal context formed by past actions on objects, coined action context. We propose TransFusion, a multimodal transformer-based architecture for short-term object interaction anticipation. Our method exploits the representational power of language by summarizing the action con-text textually, after leveraging pre-trained vision-language foundation models to extract the action context from past video frames. The summarized action context and the last observed video frame are processed by the multimodal fusion module to forecast the next object interaction. Experiments on the Ego4D next active object interaction dataset show the effectiveness of our multimodal fusion model and highlight the benefits of using the power of foundation models and language-based context summaries in a task where vision may appear to suffice. Our novel approach outperforms all state-of-the-art methods on both versions of the Ego4D dataset. A project video and code are available at https://eth-ait.github.io/transfusion-proj/.

Eye-tracking applications that utilize the human gaze in video understanding tasks have become increasingly important. To effectively automate the process of video analysis based on eye-tracking data, it is important to accurately replicate human gaze behavior. However, this task presents significant challenges due to the inherent complexity and ambiguity of human gaze patterns. In this work, we introduce a novel method for simulating human gaze behavior. Our approach uses a transformer-based reinforcement learning algorithm to train an agent that acts as a human observer, with the primary role of watching videos and simulating human gaze behavior. We employed an eye-tracking dataset gathered from videos generated by the VirtualHome simulator, with a primary focus on activity recognition. Our experimental results demonstrate the effectiveness of our gaze prediction method by highlighting its capability to replicate human gaze behavior and its applicability for downstream tasks where real human-gaze is used as input.

Humans utilize their gaze to concentrate on essential information while perceiving and interpreting intentions in videos. Incorporating human gaze into computational algorithms can significantly enhance model performance in video understanding tasks. In this work, we address a challenging and innovative task in video understanding: predicting the actions of an agent in a video based on a partial video. We introduce the Gaze-guided Action Anticipation algorithm, which establishes a visual-semantic graph from the video input. Our method utilizes a Graph Neural Network to recognize the agent’s intention and predict the action sequence to fulfill this intention. To assess the efficiency of our approach, we collect a dataset containing household activities generated in the VirtualHome environment, accompanied by human gaze data of viewing videos. Our method outperforms state-of-the-art techniques, achieving a 7% improvement in accuracy for 18-class intention recognition. This highlights the efficiency of our method in learning important features from human gaze data.

While deep neural networks are proven to be effective learning systems, their analysis is complex due to the high-dimensionality of their weight space. Persistent topological properties can be used as an additional descriptor, providing insights on how the network weights evolve during training. In this paper, we focus on convolutional neural networks, and define the topology of the space, populated by convolutional filters (i.e., kernels). We perform an extensive analysis of topological properties of the convolutional filters. Specifically, we define a metric based on persistent homology, namely, Convolutional Topology Representation, to determine an important factor in neural networks training: the generalizability of the model to the test set. We further analyse how various training methods affect the topology of convolutional layers.

We propose GazeNeRF, a 3D-aware method for the task of gaze redirection. Existing gaze redirection methods operate on 2D images and struggle to generate 3D consistent results. Instead, we build on the intuition that the face region and eyeballs are separate 3D structures that move in a coordinated yet independent fashion. Our method leverages recent advancements in conditional image-based neural radiance fields and proposes a two-stream architecture that predicts volumetric features for the face and eye regions separately. Rigidly transforming the eye features via a 3D rotation matrix provides fine-grained control over the desired gaze angle. The final, redirected image is then attained via differentiable volume compositing. Our experiments show that this architecture outperforms naively conditioned NeRF baselines as well as previous state-of-the-art 2D gaze redirection methods in terms of redirection accuracy and identity preservation. Code and models will be released for research purposes.

We present a method for inferring diverse 3D models of human-object interactions from images. Reasoning about how humans interact with objects in complex scenes from a single 2D image is a challenging task given ambiguities arising from the loss of information through projection. In addition, modeling 3D interactions requires the generalization ability towards diverse object categories and interaction types. We propose an action-conditioned modeling of interactions that allows us to infer diverse 3D arrangements of humans and objects without supervision on contact regions or 3D scene geometry. Our method extracts high-level commonsense knowledge from large language models (such as GPT-3), and applies them to perform 3D reasoning of human-object interactions. Our key insight is priors extracted from large language models can help in reasoning about human-object contacts from textural prompts only. We quantitatively evaluate the inferred 3D models on a large human-object interaction dataset and show how our method leads to better 3D reconstructions. We further qualitatively evaluate the effectiveness of our method on real images and demonstrate its generalizability towards interaction types and object categories.

Encouraged by the success of contrastive learning on image classification tasks, we propose a new self-supervised method for the structured regression task of 3D hand pose estimation. Contrastive learning makes use of unlabeled data for the purpose of representation learning via a loss formulation that encourages the learned feature representations to be invariant under any image transformation. For 3D hand pose estimation, it too is desirable to have invariance to appearance transformation such as color jitter. However, the task requires equivariance under affine transformations, such as rotation and translation. To address this issue, we propose an equivariant contrastive objective and demonstrate its effectiveness in the context of 3D hand pose estimation. We experimentally investigate the impact of invariant and equivariant contrastive objectives and show that learning equivariant features leads to better representations for the task of 3D hand pose estimation. Furthermore, we show that standard ResNets with sufficient depth, trained on additional unlabeled data, attain improvements of up to 14.5% in PA-EPE on FreiHAND and thus achieves state-of-the-art performance without any task specific, specialized architectures. Code and models are available at https://ait.ethz.ch/projects/2021/PeCLR/

It has long been hypothesized that perceptual ambiguities play an important role in aesthetic experience: A work with some ambiguity engages a viewer more than one that does not. However, current frameworks for testing this theory are limited by the availability of stimuli and data collection methods. This article presents an approach to measuring the perceptual ambiguity of a collection of images. Crowdworkers are asked to describe image content, after different viewing durations. Experiments are performed using images created with Generative Adversarial Networks, using the Artbreeder website. We show that text processing of viewer responses can provide a fine-grained way to measure and describe image ambiguities.

Humans involuntarily move their eyes when retrieving an image from memory. This motion is often similar to actually observing the image. We suggest to exploit this behavior as a new modality in human computer interaction, using the motion of the eyes as a descriptor of the image. Interaction requires the user's eyes to be tracked but no voluntary physical activity. We perform a controlled experiment and develop matching techniques using machine learning to investigate if images can be discriminated based on the gaze patterns recorded while users merely think about image. Our results indicate that image retrieval is possible with an accuracy significantly above chance. We also show that this result generalizes to images not used during training of the classifier and extends to uncontrolled settings in a realistic scenario.

Head-mounted displays cause discomfort. This is commonly attributed to conflicting depth cues, most prominently between vergence, which is consistent with object depth, and accommodation, which is adjusted to the near eye displays. It is possible to adjust the camera parameters, specifically interocular distance and vergence angles, for rendering the virtual environment to minimize this conflict. This requires dynamic adjustment of the parameters based on object depth. In an experiment based on a visual search task, we evaluate how dynamic adjustment affects visual comfort compared to fixed camera parameters. We collect objective as well as subjective data. Results show that dynamic adjustment decreases common objective measures of visual comfort such as pupil diameter and blink rate by a statistically significant margin. The subjective evaluation of categories such as fatigue or eye irritation shows a similar trend but was inconclusive. This suggests that rendering with fixed camera parameters is the better choice for head-mounted displays, at least in scenarios similar to the ones used here.

We provide the first large dataset of human fixations on physical 3D objects presented in varying viewing conditions and made of different materials. Our experimental setup is carefully designed to allow for accurate calibration and measurement. We estimate a mapping from the pair of pupil positions to 3D coordinates in space and register the presented shape with the eye tracking setup. By modeling the fixated positions on 3D shapes as a probability distribution, we analysis the similarities among different conditions. The resulting data indicates that salient features depend on the viewing direction. Stable features across different viewing directions seem to be connected to semantically meaningful parts. We also show that it is possible to estimate the gaze density maps from view dependent data. The dataset provides the necessary ground truth data for computational models of human perception in 3D.

Segmentation of Synthetic Aperture Radar (SAR) images is often only understood as the partitioning of the image into rather small regions which are homogeneous with respect to scattering processes. This paper proposes an adaption of the graph-cut image segmentation framework to the unique characteristics of polarimetric SAR images by using a Wishart-distribution based distance measure for local segmentation cues and simple, real-valued features derived from the complex-valued coherency matrix. The proposed method is evaluated on different polarimetric SAR images, for different objects of interest, and with a wide range of parameters. The results show that the proposed framework is able to derive accurate object/non-object segmentations. Best results are obtained for forest areas by usage of a log-transform of the polarimetric intensities.

Depth image-based rendering (DIBR) techniques are advanced tools in 3-D video (3DV) applications that are used to synthesize a number of additional views in a multiview-video-plus-depth (MVD) representation. The MVD format consists of video and depth sequences for a limited number of original camera views of the same scene. An inherent problem of the view synthesis concept is given by image information that is occluded in the original views and becomes visible, in the extrapolated views. To handle these disoc-clussions, we propose a DIBR algorithm with advanced inpainting methods. Our renderer enhances visual experience by taking spatial and temporal texture consistency problems into account. In order to compensate the global motion in a sequence, image registration is incorporated into the framework. The proposed method shows objective and subjective gains compared to the state-of-the-art.

Depth image-based rendering (DIBR) techniques allow for a wide variety of 3-D applications, including synthesizing additional virtual views in a multiview-video-plus-depth (MVD) representation. The MVD format consists of scene texture and depth information for a limited number of original views of the same scene. One of the main obstacles in the DIBR technique lies in the disocclusion problem which results from the fact that a scene can only be observed from a set of original views. This can lead to missing information in the generated virtual views, especially in extrapolation scenarios. Our work describes a novel algorithm that synthesizes such disoccluded textures. The proposed synthesizer enhances the visual experience by taking spatial and temporal video information into account. In order to compensate for global motion in sequences, image registration is incorporated into the framework. Objective and subjective gains are shown compared to three state-of-the-art approaches.