Felix Heide

Martin Hahner, Christos Sakaridis, Mario Bijelic et al. · eth-zurich

3D object detection is a central task for applications such as autonomous driving, in which the system needs to localize and classify surrounding traffic agents, even in the presence of adverse weather. In this paper, we address the problem of LiDAR-based 3D object detection under snowfall. Due to the difficulty of collecting and annotating training data in this setting, we propose a physically based method to simulate the effect of snowfall on real clear-weather LiDAR point clouds. Our method samples snow particles in 2D space for each LiDAR line and uses the induced geometry to modify the measurement for each LiDAR beam accordingly. Moreover, as snowfall often causes wetness on the ground, we also simulate ground wetness on LiDAR point clouds. We use our simulation to generate partially synthetic snowy LiDAR data and leverage these data for training 3D object detection models that are robust to snowfall. We conduct an extensive evaluation using several state-of-the-art 3D object detection methods and show that our simulation consistently yields significant performance gains on the real snowy STF dataset compared to clear-weather baselines and competing simulation approaches, while not sacrificing performance in clear weather. Our code is available at www.github.com/SysCV/LiDAR_snow_sim.

Gene Chou, Ilya Chugunov, Felix Heide

We investigate the generalization capabilities of neural signed distance functions (SDFs) for learning 3D object representations for unseen and unlabeled point clouds. Existing methods can fit SDFs to a handful of object classes and boast fine detail or fast inference speeds, but do not generalize well to unseen shapes. We introduce a two-stage semi-supervised meta-learning approach that transfers shape priors from labeled to unlabeled data to reconstruct unseen object categories. The first stage uses an episodic training scheme to simulate training on unlabeled data and meta-learns initial shape priors. The second stage then introduces unlabeled data with disjoint classes in a semi-supervised scheme to diversify these priors and achieve generalization. We assess our method on both synthetic data and real collected point clouds. Experimental results and analysis validate that our approach outperforms existing neural SDF methods and is capable of robust zero-shot inference on 100+ unseen classes. Code can be found at https://github.com/princeton-computational-imaging/gensdf.

Qiang Zhang, Seung-Hwan Baek, Szymon Rusinkiewicz et al.

We propose a differentiable rendering algorithm for efficient novel view synthesis. By departing from volume-based representations in favor of a learned point representation, we improve on existing methods more than an order of magnitude in memory and runtime, both in training and inference. The method begins with a uniformly-sampled random point cloud and learns per-point position and view-dependent appearance, using a differentiable splat-based renderer to evolve the model to match a set of input images. Our method is up to 300x faster than NeRF in both training and inference, with only a marginal sacrifice in quality, while using less than 10~MB of memory for a static scene. For dynamic scenes, our method trains two orders of magnitude faster than STNeRF and renders at near interactive rate, while maintaining high image quality and temporal coherence even without imposing any temporal-coherency regularizers.

Gene Chou, Yuval Bahat, Felix Heide

Probabilistic diffusion models have achieved state-of-the-art results for image synthesis, inpainting, and text-to-image tasks. However, they are still in the early stages of generating complex 3D shapes. This work proposes Diffusion-SDF, a generative model for shape completion, single-view reconstruction, and reconstruction of real-scanned point clouds. We use neural signed distance functions (SDFs) as our 3D representation to parameterize the geometry of various signals (e.g., point clouds, 2D images) through neural networks. Neural SDFs are implicit functions and diffusing them amounts to learning the reversal of their neural network weights, which we solve using a custom modulation module. Extensive experiments show that our method is capable of both realistic unconditional generation and conditional generation from partial inputs. This work expands the domain of diffusion models from learning 2D, explicit representations, to 3D, implicit representations.

Linwei Chen, Ying Fu, Kaixuan Wei et al.

Existing instance segmentation techniques are primarily tailored for high-visibility inputs, but their performance significantly deteriorates in extremely low-light environments. In this work, we take a deep look at instance segmentation in the dark and introduce several techniques that substantially boost the low-light inference accuracy. The proposed method is motivated by the observation that noise in low-light images introduces high-frequency disturbances to the feature maps of neural networks, thereby significantly degrading performance. To suppress this ``feature noise", we propose a novel learning method that relies on an adaptive weighted downsampling layer, a smooth-oriented convolutional block, and disturbance suppression learning. These components effectively reduce feature noise during downsampling and convolution operations, enabling the model to learn disturbance-invariant features. Furthermore, we discover that high-bit-depth RAW images can better preserve richer scene information in low-light conditions compared to typical camera sRGB outputs, thus supporting the use of RAW-input algorithms. Our analysis indicates that high bit-depth can be critical for low-light instance segmentation. To mitigate the scarcity of annotated RAW datasets, we leverage a low-light RAW synthetic pipeline to generate realistic low-light data. In addition, to facilitate further research in this direction, we capture a real-world low-light instance segmentation dataset comprising over two thousand paired low/normal-light images with instance-level pixel-wise annotations. Remarkably, without any image preprocessing, we achieve satisfactory performance on instance segmentation in very low light (4~\% AP higher than state-of-the-art competitors), meanwhile opening new opportunities for future research.

Praneeth Chakravarthula, Jipeng Sun, Xiao Li et al.

Today's commodity camera systems rely on compound optics to map light originating from the scene to positions on the sensor where it gets recorded as an image. To record images without optical aberrations, i.e., deviations from Gauss' linear model of optics, typical lens systems introduce increasingly complex stacks of optical elements which are responsible for the height of existing commodity cameras. In this work, we investigate flat nanophotonic computational cameras as an alternative that employs an array of skewed lenslets and a learned reconstruction approach. The optical array is embedded on a metasurface that, at 700~nm height, is flat and sits on the sensor cover glass at 2.5~mm focal distance from the sensor. To tackle the highly chromatic response of a metasurface and design the array over the entire sensor, we propose a differentiable optimization method that continuously samples over the visible spectrum and factorizes the optical modulation for different incident fields into individual lenses. We reconstruct a megapixel image from our flat imager with a learned probabilistic reconstruction method that employs a generative diffusion model to sample an implicit prior. To tackle scene-dependent aberrations in broadband, we propose a method for acquiring paired captured training data in varying illumination conditions. We assess the proposed flat camera design in simulation and with an experimental prototype, validating that the method is capable of recovering images from diverse scenes in broadband with a single nanophotonic layer.

Kaixuan Wei, Xiao Li, Johannes Froech et al.

The explosive growth of computation and energy cost of artificial intelligence has spurred strong interests in new computing modalities as potential alternatives to conventional electronic processors. Photonic processors that execute operations using photons instead of electrons, have promised to enable optical neural networks with ultra-low latency and power consumption. However, existing optical neural networks, limited by the underlying network designs, have achieved image recognition accuracy far below that of state-of-the-art electronic neural networks. In this work, we close this gap by embedding massively parallelized optical computation into flat camera optics that perform neural network computation during the capture, before recording an image on the sensor. Specifically, we harness large kernels and propose a large-kernel spatially-varying convolutional neural network learned via low-dimensional reparameterization techniques. We experimentally instantiate the network with a flat meta-optical system that encompasses an array of nanophotonic structures designed to induce angle-dependent responses. Combined with an extremely lightweight electronic backend with approximately 2K parameters we demonstrate a reconfigurable nanophotonic neural network reaches 72.76\% blind test classification accuracy on CIFAR-10 dataset, and, as such, the first time, an optical neural network outperforms the first modern digital neural network -- AlexNet (72.64\%) with 57M parameters, bringing optical neural network into modern deep learning era.

Geoffroi Côté, Fahim Mannan, Simon Thibault et al.

Most camera lens systems are designed in isolation, separately from downstream computer vision methods. Recently, joint optimization approaches that design lenses alongside other components of the image acquisition and processing pipeline -- notably, downstream neural networks -- have achieved improved imaging quality or better performance on vision tasks. However, these existing methods optimize only a subset of lens parameters and cannot optimize glass materials given their categorical nature. In this work, we develop a differentiable spherical lens simulation model that accurately captures geometrical aberrations. We propose an optimization strategy to address the challenges of lens design -- notorious for non-convex loss function landscapes and many manufacturing constraints -- that are exacerbated in joint optimization tasks. Specifically, we introduce quantized continuous glass variables to facilitate the optimization and selection of glass materials in an end-to-end design context, and couple this with carefully designed constraints to support manufacturability. In automotive object detection, we report improved detection performance over existing designs even when simplifying designs to two- or three-element lenses, despite significantly degrading the image quality.

Haiwei Zhang, Jiqing Zhang, Bo Dong et al.

We introduce a wearable single-eye emotion recognition device and a real-time approach to recognizing emotions from partial observations of an emotion that is robust to changes in lighting conditions. At the heart of our method is a bio-inspired event-based camera setup and a newly designed lightweight Spiking Eye Emotion Network (SEEN). Compared to conventional cameras, event-based cameras offer a higher dynamic range (up to 140 dB vs. 80 dB) and a higher temporal resolution. Thus, the captured events can encode rich temporal cues under challenging lighting conditions. However, these events lack texture information, posing problems in decoding temporal information effectively. SEEN tackles this issue from two different perspectives. First, we adopt convolutional spiking layers to take advantage of the spiking neural network's ability to decode pertinent temporal information. Second, SEEN learns to extract essential spatial cues from corresponding intensity frames and leverages a novel weight-copy scheme to convey spatial attention to the convolutional spiking layers during training and inference. We extensively validate and demonstrate the effectiveness of our approach on a specially collected Single-eye Event-based Emotion (SEE) dataset. To the best of our knowledge, our method is the first eye-based emotion recognition method that leverages event-based cameras and spiking neural network.

Florian Schiffers, Praneeth Chakravarthula, Nathan Matsuda et al.

The Visual Turing Test is the ultimate goal to evaluate the realism of holographic displays. Previous studies have focused on addressing challenges such as limited étendue and image quality over a large focal volume, but they have not investigated the effect of pupil sampling on the viewing experience in full 3D holograms. In this work, we tackle this problem with a novel hologram generation algorithm motivated by matching the projection operators of incoherent Light Field and coherent Wigner Function light transport. To this end, we supervise hologram computation using synthesized photographs, which are rendered on-the-fly using Light Field refocusing from stochastically sampled pupil states during optimization. The proposed method produces holograms with correct parallax and focus cues, which are important for passing the Visual Turing Test. We validate that our approach compares favorably to state-of-the-art CGH algorithms that use Light Field and Focal Stack supervision. Our experiments demonstrate that our algorithm significantly improves the realism of the viewing experience for a variety of different pupil states.

Ilya Chugunov, Yuxuan Zhang, Felix Heide

Modern mobile burst photography pipelines capture and merge a short sequence of frames to recover an enhanced image, but often disregard the 3D nature of the scene they capture, treating pixel motion between images as a 2D aggregation problem. We show that in a ''long-burst'', forty-two 12-megapixel RAW frames captured in a two-second sequence, there is enough parallax information from natural hand tremor alone to recover high-quality scene depth. To this end, we devise a test-time optimization approach that fits a neural RGB-D representation to long-burst data and simultaneously estimates scene depth and camera motion. Our plane plus depth model is trained end-to-end, and performs coarse-to-fine refinement by controlling which multi-resolution volume features the network has access to at what time during training. We validate the method experimentally, and demonstrate geometrically accurate depth reconstructions with no additional hardware or separate data pre-processing and pose-estimation steps.

Baiang Li, Ruyu Yan, Ethan Tseng et al.

While the human eye can perceive an impressive twenty stops of dynamic range, smartphone camera sensors remain limited to about twelve stops despite decades of research. A variety of high dynamic range (HDR) image capture and processing techniques have been proposed, and, in practice, they can extend the dynamic range by 3-5 stops for handheld photography. This paper proposes an approach that robustly captures dynamic range using a handheld smartphone camera and lightweight networks suitable for running on mobile devices. Our method operates indirectly on linear raw pixels in bracketed exposures. Every pixel in the final HDR image is a convex combination of input pixels in the neighborhood, adjusted for exposure, and thus avoids hallucination artifacts typical of recent deep image synthesis networks. We validate our system on both synthetic imagery and unseen real bracketed images -- we confirm zero-shot generalization of the method to smartphone camera captures. Our iterative inference architecture is capable of processing an arbitrary number of bracketed input photos, and we show examples from capture stacks containing 3--9 images. Our training process relies only on synthetic captures yet generalizes to unseen real photos from several cameras. Moreover, we show that this training scheme improves other SOTA methods over their pretrained counterparts.

Yiming Zuo, Hongyu Wen, Venkat Subramanian et al.

Depth from Defocus (DfD) is the task of estimating a dense metric depth map from a focus stack. Unlike previous works overfitting to a certain dataset, this paper focuses on the challenging and practical setting of zero-shot generalization. We first propose a new real-world DfD benchmark ZEDD, which contains 8.3x more scenes and significantly higher quality images and ground-truth depth maps compared to previous benchmarks. We also design a novel network architecture named FOSSA. FOSSA is a Transformer-based architecture with novel designs tailored to the DfD task. The key contribution is a stack attention layer with a focus distance embedding, allowing efficient information exchange across the focus stack. Finally, we develop a new training data pipeline allowing us to utilize existing large-scale RGBD datasets to generate synthetic focus stacks. Experiment results on ZEDD and other benchmarks show a significant improvement over the baselines, reducing errors by up to 55.7%. The ZEDD benchmark is released at https://zedd.cs.princeton.edu. The code and checkpoints are released at https://github.com/princeton-vl/FOSSA.

Yuval Bahat, Yuxuan Zhang, Hendrik Sommerhoff et al.

Neural volumetric representations have become a widely adopted model for radiance fields in 3D scenes. These representations are fully implicit or hybrid function approximators of the instantaneous volumetric radiance in a scene, which are typically learned from multi-view captures of the scene. We investigate the new task of neural volume super-resolution - rendering high-resolution views corresponding to a scene captured at low resolution. To this end, we propose a neural super-resolution network that operates directly on the volumetric representation of the scene. This approach allows us to exploit an advantage of operating in the volumetric domain, namely the ability to guarantee consistent super-resolution across different viewing directions. To realize our method, we devise a novel 3D representation that hinges on multiple 2D feature planes. This allows us to super-resolve the 3D scene representation by applying 2D convolutional networks on the 2D feature planes. We validate the proposed method by super-resolving multi-view consistent views on a diverse set of unseen 3D scenes, confirming qualitative and quantitatively favorable quality over existing approaches.

Hannah Dröge, Zorah Lähner, Yuval Bahat et al.

Permutation matrices play a key role in matching and assignment problems across the fields, especially in computer vision and robotics. However, memory for explicitly representing permutation matrices grows quadratically with the size of the problem, prohibiting large problem instances. In this work, we propose to tackle the curse of dimensionality of large permutation matrices by approximating them using low-rank matrix factorization, followed by a nonlinearity. To this end, we rely on the Kissing number theory to infer the minimal rank required for representing a permutation matrix of a given size, which is significantly smaller than the problem size. This leads to a drastic reduction in computation and memory costs, e.g., up to $3$ orders of magnitude less memory for a problem of size $n=20000$, represented using $8.4\times10^5$ elements in two small matrices instead of using a single huge matrix with $4\times 10^8$ elements. The proposed representation allows for accurate representations of large permutation matrices, which in turn enables handling large problems that would have been infeasible otherwise. We demonstrate the applicability and merits of the proposed approach through a series of experiments on a range of problems that involve predicting permutation matrices, from linear and quadratic assignment to shape matching problems.

Lili Gao, Yanbo Xu, William Koch et al.

We introduce ScenarioControl, the first vision-language control mechanism for learned driving scenario generation. Given a text prompt or an input image, Scenario-Control synthesizes diverse, realistic 3D scenario rollouts - including map, 3D boxes of reactive actors over time, pedestrians, driving infrastructure, and ego camera observations. The method generates scenes in a vectorized latent space that represents road structure and dynamic agents jointly. To connect multimodal control with sparse vectorized scene elements, we propose a cross-global control mechanism that integrates crossattention with a lightweight global-context branch, enabling fine-grained control over road layout and traffic conditions while preserving realism. The method produces temporally consistent scenario rollouts from the perspectives different actors in the scene, supporting long-horizon continuation of driving scenarios. To facilitate training and evaluation, we release a dataset with text annotations aligned to vectorized map structures. Extensive experiments validate that the control adherence and fidelity of ScenarioControl compare favorable to all tested methods across all experiments. Project webpage: https://light.princeton.edu/ScenarioControl

Parker Ewen, Dmitriy Rivkin, Mario Bijelic et al.

Autonomous highway driving, especially for long-haul heavy trucks, requires detecting objects at long ranges beyond 500 meters to satisfy braking distance requirements at high speeds. At long distances, vehicles and other critical objects occupy only a few pixels in high-resolution images, causing state-of-the-art object detectors to fail. This challenge is compounded by the limited effective range of commercially available LiDAR sensors, which fall short of ultra-long range thresholds because of quadratic loss of resolution with distance, making image-based detection the most practically scalable solution given commercially available sensor constraints. We introduce Telescope, a two-stage detection model designed for ultra-long range autonomous driving. Alongside a powerful detection backbone, this model contains a novel re-sampling layer and image transformation to address the fundamental challenges of detecting small, distant objects. Telescope achieves $76\%$ relative improvement in mAP in ultra-long range detection compared to state-of-the-art methods (improving from an absolute mAP of 0.185 to 0.326 at distances beyond 250 meters), requires minimal computational overhead, and maintains strong performance across all detection ranges.

Filippo Ghilotti, Edoardo Palladin, Samuel Brucker et al.

Safe highway autonomy for heavy trucks remains an open and unsolved challenge: due to long braking distances, scene understanding of hundreds of meters is required for anticipatory planning and to allow safe braking margins. However, existing driving datasets primarily cover urban scenes, with perception effectively limited to short ranges of only up to 100 meters. To address this gap, we introduce TruckDrive, a highway-scale multimodal driving dataset, captured with a sensor suite purpose-built for long range sensing: seven long-range FMCW LiDARs measuring range and radial velocity, three high-resolution short-range LiDARs, eleven 8MP surround cameras with varying focal lengths and ten 4D FMCW radars. The dataset offers 475 thousands samples with 165 thousands densely annotated frames for driving perception benchmarking up to 1,000 meters for 2D detection and 400 meters for 3D detection, depth estimation, tracking, planning and end to end driving over 20 seconds sequences at highway speeds. We find that state-of-the-art autonomous driving models do not generalize to ranges beyond 150 meters, with drops between 31% and 99% in 3D perception tasks, exposing a systematic long-range gap that current architectures and training signals cannot close.

Dmitriy Rivkin, Parker Ewen, Lili Gao et al.

Recent video diffusion models achieve high-quality generation through recurrent frame processing where each frame generation depends on previous frames. However, this recurrent mechanism means that training such models in the pixel domain incurs prohibitive memory costs, as activations accumulate across the entire video sequence. This fundamental limitation also makes fine-tuning these models with pixel-wise losses computationally intractable for long or high-resolution videos. This paper introduces ChopGrad, a truncated backpropagation scheme for video decoding, limiting gradient computation to local frame windows while maintaining global consistency. We provide a theoretical analysis of this approximation and show that it enables efficient fine-tuning with frame-wise losses. ChopGrad reduces training memory from scaling linearly with the number of video frames (full backpropagation) to constant memory, and compares favorably to existing state-of-the-art video diffusion models across a suite of conditional video generation tasks with pixel-wise losses, including video super-resolution, video inpainting, video enhancement of neural-rendered scenes, and controlled driving video generation.

Ilya Chugunov, Amogh Joshi, Kiran Murthy et al.

Challenging to capture, and challenging to display on a cellphone screen, the panorama paradoxically remains both a staple and underused feature of modern mobile camera applications. In this work we address both of these challenges with a spherical neural light field model for implicit panoramic image stitching and re-rendering; able to accommodate for depth parallax, view-dependent lighting, and local scene motion and color changes during capture. Fit during test-time to an arbitrary path panoramic video capture -- vertical, horizontal, random-walk -- these neural light spheres jointly estimate the camera path and a high-resolution scene reconstruction to produce novel wide field-of-view projections of the environment. Our single-layer model avoids expensive volumetric sampling, and decomposes the scene into compact view-dependent ray offset and color components, with a total model size of 80 MB per scene, and real-time (50 FPS) rendering at 1080p resolution. We demonstrate improved reconstruction quality over traditional image stitching and radiance field methods, with significantly higher tolerance to scene motion and non-ideal capture settings.

Suhyun Shin, Seokjun Choi, Felix Heide et al.

Hyperspectral 3D imaging aims to acquire both depth and spectral information of a scene. However, existing methods are either prohibitively expensive and bulky or compromise on spectral and depth accuracy. In this work, we present Dispersed Structured Light (DSL), a cost-effective and compact method for accurate hyperspectral 3D imaging. DSL modifies a traditional projector-camera system by placing a sub-millimeter thick diffraction grating film front of the projector. The grating disperses structured light based on light wavelength. To utilize the dispersed structured light, we devise a model for dispersive projection image formation and a per-pixel hyperspectral 3D reconstruction method. We validate DSL by instantiating a compact experimental prototype. DSL achieves spectral accuracy of 18.8nm full-width half-maximum (FWHM) and depth error of 1mm. We demonstrate that DSL outperforms prior work on practical hyperspectral 3D imaging. DSL promises accurate and practical hyperspectral 3D imaging for diverse application domains, including computer vision and graphics, cultural heritage, geology, and biology.

Yujin Jeon, Eunsue Choi, Youngchan Kim et al.

Image datasets are essential not only in validating existing methods in computer vision but also in developing new methods. Most existing image datasets focus on trichromatic intensity images to mimic human vision. However, polarization and spectrum, the wave properties of light that animals in harsh environments and with limited brain capacity often rely on, remain underrepresented in existing datasets. Although spectro-polarimetric datasets exist, these datasets have insufficient object diversity, limited illumination conditions, linear-only polarization data, and inadequate image count. Here, we introduce two spectro-polarimetric datasets: trichromatic Stokes images and hyperspectral Stokes images. These novel datasets encompass both linear and circular polarization; they introduce multiple spectral channels; and they feature a broad selection of real-world scenes. With our dataset in hand, we analyze the spectro-polarimetric image statistics, develop efficient representations of such high-dimensional data, and evaluate spectral dependency of shape-from-polarization methods. As such, the proposed dataset promises a foundation for data-driven spectro-polarimetric imaging and vision research. Dataset and code will be publicly available.

Filippo Ghilotti, Samuel Brucker, Nahku Saidy et al.

Unlabeled LiDAR logs, in autonomous driving applications, are inherently a gold mine of dense 3D geometry hiding in plain sight - yet they are almost useless without human labels, highlighting a dominant cost barrier for autonomous-perception research. In this work we tackle this bottleneck by leveraging temporal-geometric consistency across LiDAR sweeps to lift and fuse cues from text and 2D vision foundation models directly into 3D, without any manual input. We introduce an unsupervised multi-modal pseudo-labeling method relying on strong geometric priors learned from temporally accumulated LiDAR maps, alongside with a novel iterative update rule that enforces joint geometric-semantic consistency, and vice-versa detecting moving objects from inconsistencies. Our method simultaneously produces 3D semantic labels, 3D bounding boxes, and dense LiDAR scans, demonstrating robust generalization across three datasets. We experimentally validate that our method compares favorably to existing semantic segmentation and object detection pseudo-labeling methods, which often require additional manual supervision. We confirm that even a small fraction of our geometrically consistent, densified LiDAR improves depth prediction by 51.5% and 22.0% MAE in the 80-150 and 150-250 meters range, respectively.

Amogh Joshi, Julian Ost, Felix Heide

Unbounded 3D world generation is emerging as a foundational task for scene modeling in computer vision, graphics, and robotics. In this work, we present WorldFlow3D, a novel method capable of generating unbounded 3D worlds. Building upon a foundational property of flow matching - namely, defining a path of transport between two data distributions - we model 3D generation more generally as a problem of flowing through 3D data distributions, not limited to conditional denoising. We find that our latent-free flow approach generates causal and accurate 3D structure, and can use this as an intermediate distribution to guide the generation of more complex structure and high-quality texture - all while converging more rapidly than existing methods. We enable controllability over generated scenes with vectorized scene layout conditions for geometric structure control and visual texture control through scene attributes. We confirm the effectiveness of WorldFlow3D on both real outdoor driving scenes and synthetic indoor scenes, validating cross-domain generalizability and high-quality generation on real data distributions. We confirm favorable scene generation fidelity over approaches in all tested settings for unbounded scene generation. For more, see https://light.princeton.edu/worldflow3d.

Baiang Li, Wenhao Chai, Felix Heide

Large-scale visual learning is increasingly limited by training cost. Existing knowledge distillation methods transfer from a stronger teacher to a weaker student for compression or final-accuracy improvement. We instead investigate distillation to accelerate the training of strong students. We propose a generalizable plug-and-play recipe that freezes a weaker teacher, applies distillation only in early training, and turns it off once the student reaches and surpasses teacher-level performance. For ImageNet and CIFAR classification, this strategy reaches target thresholds much earlier, with up to 4.8 times speedup measured by epochs. We confirm that the method generalizes to other tasks and report 1.7 times epoch speedup for object detection on the COCO dataset, and 2.5 times earlier target-FID crossing for diffusion generation on the CIFAR-10 dataset, measured in steps. These findings validate our method as a universal speedup mechanism for visual learning.

Dominik Scheuble, Chenyang Lei, Seung-Hwan Baek et al.

Lidar has become a cornerstone sensing modality for 3D vision, especially for large outdoor scenarios and autonomous driving. Conventional lidar sensors are capable of providing centimeter-accurate distance information by emitting laser pulses into a scene and measuring the time-of-flight (ToF) of the reflection. However, the polarization of the received light that depends on the surface orientation and material properties is usually not considered. As such, the polarization modality has the potential to improve scene reconstruction beyond distance measurements. In this work, we introduce a novel long-range polarization wavefront lidar sensor (PolLidar) that modulates the polarization of the emitted and received light. Departing from conventional lidar sensors, PolLidar allows access to the raw time-resolved polarimetric wavefronts. We leverage polarimetric wavefronts to estimate normals, distance, and material properties in outdoor scenarios with a novel learned reconstruction method. To train and evaluate the method, we introduce a simulated and real-world long-range dataset with paired raw lidar data, ground truth distance, and normal maps. We find that the proposed method improves normal and distance reconstruction by 53\% mean angular error and 41\% mean absolute error compared to existing shape-from-polarization (SfP) and ToF methods. Code and data are open-sourced at https://light.princeton.edu/pollidar.

Chenyang Lei, Liyi Chen, Jun Cen et al.

Foundation models like ChatGPT and Sora that are trained on a huge scale of data have made a revolutionary social impact. However, it is extremely challenging for sensors in many different fields to collect similar scales of natural images to train strong foundation models. To this end, this work presents a simple and effective framework, SimCMF, to study an important problem: cross-modal fine-tuning from vision foundation models trained on natural RGB images to other imaging modalities of different physical properties (e.g., polarization). In SimCMF, we conduct a thorough analysis of different basic components from the most naive design and ultimately propose a novel cross-modal alignment module to address the modality misalignment problem. We apply SimCMF to a representative vision foundation model Segment Anything Model (SAM) to support any evaluated new imaging modality. Given the absence of relevant benchmarks, we construct a benchmark for performance evaluation. Our experiments confirm the intriguing potential of transferring vision foundation models in enhancing other sensors' performance. SimCMF can improve the segmentation performance (mIoU) from 22.15% to 53.88% on average for evaluated modalities and consistently outperforms other baselines. The code is available at https://github.com/mt-cly/SimCMF

Tobias Gruber, Mario Bijelic, Felix Heide et al.

This work introduces an evaluation benchmark for depth estimation and completion using high-resolution depth measurements with angular resolution of up to 25" (arcsecond), akin to a 50 megapixel camera with per-pixel depth available. Existing datasets, such as the KITTI benchmark, provide only sparse reference measurements with an order of magnitude lower angular resolution - these sparse measurements are treated as ground truth by existing depth estimation methods. We propose an evaluation methodology in four characteristic automotive scenarios recorded in varying weather conditions (day, night, fog, rain). As a result, our benchmark allows us to evaluate the robustness of depth sensing methods in adverse weather and different driving conditions. Using the proposed evaluation data, we demonstrate that current stereo approaches provide significantly more stable depth estimates than monocular methods and lidar completion in adverse weather. Data and code are available at https://github.com/gruberto/PixelAccurateDepthBenchmark.git.

Mario Bijelic, Tobias Gruber, Fahim Mannan et al.

The fusion of multimodal sensor streams, such as camera, lidar, and radar measurements, plays a critical role in object detection for autonomous vehicles, which base their decision making on these inputs. While existing methods exploit redundant information in good environmental conditions, they fail in adverse weather where the sensory streams can be asymmetrically distorted. These rare "edge-case" scenarios are not represented in available datasets, and existing fusion architectures are not designed to handle them. To address this challenge we present a novel multimodal dataset acquired in over 10,000km of driving in northern Europe. Although this dataset is the first large multimodal dataset in adverse weather, with 100k labels for lidar, camera, radar, and gated NIR sensors, it does not facilitate training as extreme weather is rare. To this end, we present a deep fusion network for robust fusion without a large corpus of labeled training data covering all asymmetric distortions. Departing from proposal-level fusion, we propose a single-shot model that adaptively fuses features, driven by measurement entropy. We validate the proposed method, trained on clean data, on our extensive validation dataset. Code and data are available here https://github.com/princeton-computational-imaging/SeeingThroughFog.

Tobias Gruber, Frank Julca-Aguilar, Mario Bijelic et al.

We present an imaging framework which converts three images from a gated camera into high-resolution depth maps with depth accuracy comparable to pulsed lidar measurements. Existing scanning lidar systems achieve low spatial resolution at large ranges due to mechanically-limited angular sampling rates, restricting scene understanding tasks to close-range clusters with dense sampling. Moreover, today's pulsed lidar scanners suffer from high cost, power consumption, large form-factors, and they fail in the presence of strong backscatter. We depart from point scanning and demonstrate that it is possible to turn a low-cost CMOS gated imager into a dense depth camera with at least 80m range - by learning depth from three gated images. The proposed architecture exploits semantic context across gated slices, and is trained on a synthetic discriminator loss without the need of dense depth labels. The proposed replacement for scanning lidar systems is real-time, handles back-scatter and provides dense depth at long ranges. We validate our approach in simulation and on real-world data acquired over 4,000km driving in northern Europe. Data and code are available at https://github.com/gruberto/Gated2Depth.

Chenyang Lei, Liyi Chen, Jun Cen et al.

Foundation models like ChatGPT and Sora that are trained on a huge scale of data have made a revolutionary social impact. However, it is extremely challenging for sensors in many different fields to collect similar scales of natural images to train strong foundation models. To this end, this work presents a simple and effective framework SimMAT to study an open problem: the transferability from vision foundation models trained on natural RGB images to other image modalities of different physical properties (e.g., polarization). SimMAT consists of a modality-agnostic transfer layer (MAT) and a pretrained foundation model. We apply SimMAT to a representative vision foundation model Segment Anything Model (SAM) to support any evaluated new image modality. Given the absence of relevant benchmarks, we construct a new benchmark to evaluate the transfer learning performance. Our experiments confirm the intriguing potential of transferring vision foundation models in enhancing other sensors' performance. Specifically, SimMAT can improve the segmentation performance (mIoU) from 22.15% to 53.88% on average for evaluated modalities and consistently outperforms other baselines. We hope that SimMAT can raise awareness of cross-modal transfer learning and benefit various fields for better results with vision foundation models.

Cheng Zheng, William Koch, Baiang Li et al.

Hierarchical structures of motion exist across research fields, including computer vision, graphics, and robotics, where complex dynamics typically arise from coordinated interactions among simpler motion components. Existing methods to model such dynamics typically rely on manually-defined or heuristic hierarchies with fixed motion primitives, limiting their generalizability across different tasks. In this work, we propose a general hierarchical motion modeling method that learns structured, interpretable motion relationships directly from data. Our method represents observed motions using graph-based hierarchies, explicitly decomposing global absolute motions into parent-inherited patterns and local motion residuals. We formulate hierarchy inference as a differentiable graph learning problem, where vertices represent elemental motions and directed edges capture learned parent-child dependencies through graph neural networks. We evaluate our hierarchical reconstruction approach on three examples: 1D translational motion, 2D rotational motion, and dynamic 3D scene deformation via Gaussian splatting. Experimental results show that our method reconstructs the intrinsic motion hierarchy in 1D and 2D cases, and produces more realistic and interpretable deformations compared to the baseline on dynamic 3D Gaussian splatting scenes. By providing an adaptable, data-driven hierarchical modeling paradigm, our method offers a formulation applicable to a broad range of motion-centric tasks. Project Page: https://light.princeton.edu/HEIR/

Ilya Chugunov, David Shustin, Ruyu Yan et al.

Each photo in an image burst can be considered a sample of a complex 3D scene: the product of parallax, diffuse and specular materials, scene motion, and illuminant variation. While decomposing all of these effects from a stack of misaligned images is a highly ill-conditioned task, the conventional align-and-merge burst pipeline takes the other extreme: blending them into a single image. In this work, we propose a versatile intermediate representation: a two-layer alpha-composited image plus flow model constructed with neural spline fields -- networks trained to map input coordinates to spline control points. Our method is able to, during test-time optimization, jointly fuse a burst image capture into one high-resolution reconstruction and decompose it into transmission and obstruction layers. Then, by discarding the obstruction layer, we can perform a range of tasks including seeing through occlusions, reflection suppression, and shadow removal. Validated on complex synthetic and in-the-wild captures we find that, with no post-processing steps or learned priors, our generalizable model is able to outperform existing dedicated single-image and multi-view obstruction removal approaches.

David Borts, Erich Liang, Tim Brödermann et al.

Neural fields have been broadly investigated as scene representations for the reproduction and novel generation of diverse outdoor scenes, including those autonomous vehicles and robots must handle. While successful approaches for RGB and LiDAR data exist, neural reconstruction methods for radar as a sensing modality have been largely unexplored. Operating at millimeter wavelengths, radar sensors are robust to scattering in fog and rain, and, as such, offer a complementary modality to active and passive optical sensing techniques. Moreover, existing radar sensors are highly cost-effective and deployed broadly in robots and vehicles that operate outdoors. We introduce Radar Fields - a neural scene reconstruction method designed for active radar imagers. Our approach unites an explicit, physics-informed sensor model with an implicit neural geometry and reflectance model to directly synthesize raw radar measurements and extract scene occupancy. The proposed method does not rely on volume rendering. Instead, we learn fields in Fourier frequency space, supervised with raw radar data. We validate the effectiveness of the method across diverse outdoor scenarios, including urban scenes with dense vehicles and infrastructure, and in harsh weather scenarios, where mm-wavelength sensing is especially favorable.

Luke Rowe, Roger Girgis, Anthony Gosselin et al.

Evaluating autonomous vehicle stacks (AVs) in simulation typically involves replaying driving logs from real-world recorded traffic. However, agents replayed from offline data are not reactive and hard to intuitively control. Existing approaches address these challenges by proposing methods that rely on heuristics or generative models of real-world data but these approaches either lack realism or necessitate costly iterative sampling procedures to control the generated behaviours. In this work, we take an alternative approach and propose CtRL-Sim, a method that leverages return-conditioned offline reinforcement learning (RL) to efficiently generate reactive and controllable traffic agents. Specifically, we process real-world driving data through a physics-enhanced Nocturne simulator to generate a diverse offline RL dataset, annotated with various rewards. With this dataset, we train a return-conditioned multi-agent behaviour model that allows for fine-grained manipulation of agent behaviours by modifying the desired returns for the various reward components. This capability enables the generation of a wide range of driving behaviours beyond the scope of the initial dataset, including adversarial behaviours. We show that CtRL-Sim can generate realistic safety-critical scenarios while providing fine-grained control over agent behaviours.

Kei Ikemura, Yiming Huang, Felix Heide et al.

Existing depth sensors are imperfect and may provide inaccurate depth values in challenging scenarios, such as in the presence of transparent or reflective objects. In this work, we present a general framework that leverages polarization imaging to improve inaccurate depth measurements from various depth sensors. Previous polarization-based depth enhancement methods focus on utilizing pure physics-based formulas for a single sensor. In contrast, our method first adopts a learning-based strategy where a neural network is trained to estimate a dense and complete depth map from polarization data and a sensor depth map from different sensors. To further improve the performance, we propose a Polarization Prompt Fusion Tuning (PPFT) strategy to effectively utilize RGB-based models pre-trained on large-scale datasets, as the size of the polarization dataset is limited to train a strong model from scratch. We conducted extensive experiments on a public dataset, and the results demonstrate that the proposed method performs favorably compared to existing depth enhancement baselines. Code and demos are available at https://lastbasket.github.io/PPFT/.

Luke Rowe, Roger Girgis, Anthony Gosselin et al.

We introduce Scenario Dreamer, a fully data-driven generative simulator for autonomous vehicle planning that generates both the initial traffic scene - comprising a lane graph and agent bounding boxes - and closed-loop agent behaviours. Existing methods for generating driving simulation environments encode the initial traffic scene as a rasterized image and, as such, require parameter-heavy networks that perform unnecessary computation due to many empty pixels in the rasterized scene. Moreover, we find that existing methods that employ rule-based agent behaviours lack diversity and realism. Scenario Dreamer instead employs a novel vectorized latent diffusion model for initial scene generation that directly operates on the vectorized scene elements and an autoregressive Transformer for data-driven agent behaviour simulation. Scenario Dreamer additionally supports scene extrapolation via diffusion inpainting, enabling the generation of unbounded simulation environments. Extensive experiments show that Scenario Dreamer outperforms existing generative simulators in realism and efficiency: the vectorized scene-generation base model achieves superior generation quality with around 2x fewer parameters, 6x lower generation latency, and 10x fewer GPU training hours compared to the strongest baseline. We confirm its practical utility by showing that reinforcement learning planning agents are more challenged in Scenario Dreamer environments than traditional non-generative simulation environments, especially on long and adversarial driving environments.

Samuel Brucker, Stefanie Walz, Mario Bijelic et al.

Gated cameras flood-illuminate a scene and capture the time-gated impulse response of a scene. By employing nanosecond-scale gates, existing sensors are capable of capturing mega-pixel gated images, delivering dense depth improving on today's LiDAR sensors in spatial resolution and depth precision. Although gated depth estimation methods deliver a million of depth estimates per frame, their resolution is still an order below existing RGB imaging methods. In this work, we combine high-resolution stereo HDR RCCB cameras with gated imaging, allowing us to exploit depth cues from active gating, multi-view RGB and multi-view NIR sensing -- multi-view and gated cues across the entire spectrum. The resulting capture system consists only of low-cost CMOS sensors and flood-illumination. We propose a novel stereo-depth estimation method that is capable of exploiting these multi-modal multi-view depth cues, including the active illumination that is measured by the RCCB camera when removing the IR-cut filter. The proposed method achieves accurate depth at long ranges, outperforming the next best existing method by 39% for ranges of 100 to 220m in MAE on accumulated LiDAR ground-truth. Our code, models and datasets are available at https://light.princeton.edu/gatedrccbstereo/ .

Edoardo Palladin, Roland Dietze, Praveen Narayanan et al.

Multimodal sensor fusion is an essential capability for autonomous robots, enabling object detection and decision-making in the presence of failing or uncertain inputs. While recent fusion methods excel in normal environmental conditions, these approaches fail in adverse weather, e.g., heavy fog, snow, or obstructions due to soiling. We introduce a novel multi-sensor fusion approach tailored to adverse weather conditions. In addition to fusing RGB and LiDAR sensors, which are employed in recent autonomous driving literature, our sensor fusion stack is also capable of learning from NIR gated camera and radar modalities to tackle low light and inclement weather. We fuse multimodal sensor data through attentive, depth-based blending schemes, with learned refinement on the Bird's Eye View (BEV) plane to combine image and range features effectively. Our detections are predicted by a transformer decoder that weighs modalities based on distance and visibility. We demonstrate that our method improves the reliability of multimodal sensor fusion in autonomous vehicles under challenging weather conditions, bridging the gap between ideal conditions and real-world edge cases. Our approach improves average precision by 17.2 AP compared to the next best method for vulnerable pedestrians in long distances and challenging foggy scenes. Our project page is available at https://light.princeton.edu/samfusion/

Jan Philipp Schneider, Pratik Singh Bisht, Ilya Chugunov et al.

Learning editable high-resolution scene representations for dynamic scenes is an open problem with applications across the domains from autonomous driving to creative editing - the most successful approaches today make a trade-off between editability and supporting scene complexity: neural atlases represent dynamic scenes as two deforming image layers, foreground and background, which are editable in 2D, but break down when multiple objects occlude and interact. In contrast, scene graph models make use of annotated data such as masks and bounding boxes from autonomous-driving datasets to capture complex 3D spatial relationships, but their implicit volumetric node representations are challenging to edit view-consistently. We propose Neural Atlas Graphs (NAGs), a hybrid high-resolution scene representation, where every graph node is a view-dependent neural atlas, facilitating both 2D appearance editing and 3D ordering and positioning of scene elements. Fit at test-time, NAGs achieve state-of-the-art quantitative results on the Waymo Open Dataset - by 5 dB PSNR increase compared to existing methods - and make environmental editing possible in high resolution and visual quality - creating counterfactual driving scenarios with new backgrounds and edited vehicle appearance. We find that the method also generalizes beyond driving scenes and compares favorably - by more than 7 dB in PSNR - to recent matting and video editing baselines on the DAVIS video dataset with a diverse set of human and animal-centric scenes. Project Page: https://princeton-computational-imaging.github.io/nag/

Julian Ost, Andrea Ramazzina, Amogh Joshi et al.

Large-scale scene data is essential for training and testing in robot learning. Neural reconstruction methods have promised the capability of reconstructing large physically-grounded outdoor scenes from captured sensor data. However, these methods have baked-in static environments and only allow for limited scene control -- they are functionally constrained in scene and trajectory diversity by the captures from which they are reconstructed. In contrast, generating driving data with recent image or video diffusion models offers control, however, at the cost of geometry grounding and causality. In this work, we aim to bridge this gap and present a method that directly generates large-scale 3D driving scenes with accurate geometry, allowing for causal novel view synthesis with object permanence and explicit 3D geometry estimation. The proposed method combines the generation of a proxy geometry and environment representation with score distillation from learned 2D image priors. We find that this approach allows for high controllability, enabling the prompt-guided geometry and high-fidelity texture and structure that can be conditioned on map layouts -- producing realistic and geometrically consistent 3D generations of complex driving scenes.

Bowen Feng, Zhiting Mei, Baiang Li et al.

While autonomous driving (AD) stacks struggle with decision making under partial observability and real-world complexity, human drivers are capable of commonsense reasoning to make near-optimal decisions with limited information. Recent work has attempted to leverage finetuned Vision-Language Models (VLMs) for trajectory planning at inference time to emulate human behavior. Despite their success in benchmark evaluations, these methods are often impractical to deploy (a 70B parameter VLM inference at merely 8 tokens per second requires more than 160G of memory), and their monolithic network structure prohibits safety decomposition. To bridge this gap, we propose VLM-Embedded Reasoning for autonomous Driving (VERDI), a training-time framework that distills the reasoning process and commonsense knowledge of VLMs into the AD stack. VERDI augments modular differentiable end-to-end (e2e) AD models by aligning intermediate module outputs at the perception, prediction, and planning stages with text features explaining the driving reasoning process produced by VLMs. By encouraging alignment in latent space, VERDI enables the modular AD stack to internalize structured reasoning, without incurring the inference-time costs of large VLMs. We demonstrate the effectiveness of our method on the NuScenes dataset and find that VERDI outperforms existing e2e methods that do not embed reasoning by 10% in $\ell_{2}$ distance, while maintaining high inference speed.

Julian Ost, Tanushree Banerjee, Mario Bijelic et al.

Today, most methods for image understanding tasks rely on feed-forward neural networks. While this approach has allowed for empirical accuracy, efficiency, and task adaptation via fine-tuning, it also comes with fundamental disadvantages. Existing networks often struggle to generalize across different datasets, even on the same task. By design, these networks ultimately reason about high-dimensional scene features, which are challenging to analyze. This is true especially when attempting to predict 3D information based on 2D images. We propose to recast 3D multi-object tracking from RGB cameras as an \emph{Inverse Rendering (IR)} problem, by optimizing via a differentiable rendering pipeline over the latent space of pre-trained 3D object representations and retrieve the latents that best represent object instances in a given input image. To this end, we optimize an image loss over generative latent spaces that inherently disentangle shape and appearance properties. We investigate not only an alternate take on tracking but our method also enables examining the generated objects, reasoning about failure situations, and resolving ambiguous cases. We validate the generalization and scaling capabilities of our method by learning the generative prior exclusively from synthetic data and assessing camera-based 3D tracking on the nuScenes and Waymo datasets. Both these datasets are completely unseen to our method and do not require fine-tuning. Videos and code are available at https://light.princeton.edu/inverse-rendering-tracking/.

Edoardo Palladin, Samuel Brucker, Filippo Ghilotti et al.

Outside of urban hubs, autonomous cars and trucks have to master driving on intercity highways. Safe, long-distance highway travel at speeds exceeding 100 km/h demands perception distances of at least 250 m, which is about five times the 50-100m typically addressed in city driving, to allow sufficient planning and braking margins. Increasing the perception ranges also allows to extend autonomy from light two-ton passenger vehicles to large-scale forty-ton trucks, which need a longer planning horizon due to their high inertia. However, most existing perception approaches focus on shorter ranges and rely on Bird's Eye View (BEV) representations, which incur quadratic increases in memory and compute costs as distance grows. To overcome this limitation, we built on top of a sparse representation and introduced an efficient 3D encoding of multi-modal and temporal features, along with a novel self-supervised pre-training scheme that enables large-scale learning from unlabeled camera-LiDAR data. Our approach extends perception distances to 250 meters and achieves an 26.6% improvement in mAP in object detection and a decrease of 30.5% in Chamfer Distance in LiDAR forecasting compared to existing methods, reaching distances up to 250 meters. Project Page: https://light.princeton.edu/lrs4fusion/

Kaixuan Wei, Hector A. Jimenez-Romero, Hadi Amata et al.

Differentiable optics, as an emerging paradigm that jointly optimizes optics and (optional) image processing algorithms, has made innovative optical designs possible across a broad range of applications. Many of these systems utilize diffractive optical components (DOEs) for holography, PSF engineering, or wavefront shaping. Existing approaches have, however, mostly remained limited to laboratory prototypes, owing to a large quality gap between simulation and manufactured devices. We aim at lifting the fundamental technical barriers to the practical use of learned diffractive optical systems. To this end, we propose a fabrication-aware design pipeline for diffractive optics fabricated by direct-write grayscale lithography followed by nano-imprinting replication, which is directly suited for inexpensive mass production of large area designs. We propose a super-resolved neural lithography model that can accurately predict the 3D geometry generated by the fabrication process. This model can be seamlessly integrated into existing differentiable optics frameworks, enabling fabrication-aware, end-to-end optimization of computational optical systems. To tackle the computational challenges, we also devise tensor-parallel compute framework centered on distributing large-scale FFT computation across many GPUs. As such, we demonstrate large scale diffractive optics designs up to 32.16 mm $\times$ 21.44 mm, simulated on grids of up to 128,640 by 85,760 feature points. We find adequate agreement between simulation and fabricated prototypes for applications such as holography and PSF engineering. We also achieve high image quality from an imaging system comprised only of a single DOE, with images processed only by a Wiener filter utilizing the simulation PSF. We believe our findings lift the fabrication limitations for real-world applications of diffractive optics and differentiable optical design.

Juhyung Choi, Jinnyeong Kim, Seokjun Choi et al.

Achieving robust stereo 3D imaging under diverse illumination conditions is an important however challenging task, due to the limited dynamic ranges (DRs) of cameras, which are significantly smaller than real world DR. As a result, the accuracy of existing stereo depth estimation methods is often compromised by under- or over-exposed images. Here, we introduce dual-exposure stereo for extended dynamic range 3D imaging. We develop automatic dual-exposure control method that adjusts the dual exposures, diverging them when the scene DR exceeds the camera DR, thereby providing information about broader DR. From the captured dual-exposure stereo images, we estimate depth using motion-aware dual-exposure stereo network. To validate our method, we develop a robot-vision system, collect stereo video datasets, and generate a synthetic dataset. Our method outperforms other exposure control methods.

Stefanie Walz, Mario Bijelic, Andrea Ramazzina et al.

We propose Gated Stereo, a high-resolution and long-range depth estimation technique that operates on active gated stereo images. Using active and high dynamic range passive captures, Gated Stereo exploits multi-view cues alongside time-of-flight intensity cues from active gating. To this end, we propose a depth estimation method with a monocular and stereo depth prediction branch which are combined in a final fusion stage. Each block is supervised through a combination of supervised and gated self-supervision losses. To facilitate training and validation, we acquire a long-range synchronized gated stereo dataset for automotive scenarios. We find that the method achieves an improvement of more than 50 % MAE compared to the next best RGB stereo method, and 74 % MAE to existing monocular gated methods for distances up to 160 m. Our code,models and datasets are available here.

Fatemeh Azimi, Fahim Mannan, Felix Heide

In this work, we study self-supervised multiple object tracking without using any video-level association labels. We propose to cast the problem of multiple object tracking as learning the frame-wise associations between detections in consecutive frames. To this end, we propose differentiable soft object assignment for object association, making it possible to learn features tailored to object association with differentiable end-to-end training. With this training approach in hand, we develop an appearance-based model for learning instance-aware object features used to construct a cost matrix based on the pairwise distances between the object features. We train our model using temporal and multi-view data, where we obtain association pseudo-labels using optical flow and disparity information. Unlike most self-supervised tracking methods that rely on pretext tasks for learning the feature correspondences, our method is directly optimized for cross-object association in complex scenarios. As such, the proposed method offers a reidentification-based MOT approach that is robust to training hyperparameters and does not suffer from local minima, which are a challenge in self-supervised methods. We evaluate our proposed model on the KITTI, Waymo, nuScenes, and Argoverse datasets, consistently improving over other unsupervised methods ($7.8\%$ improvement in association accuracy on nuScenes).

Andrea Ramazzina, Mario Bijelic, Stefanie Walz et al.

Vision in adverse weather conditions, whether it be snow, rain, or fog is challenging. In these scenarios, scattering and attenuation severly degrades image quality. Handling such inclement weather conditions, however, is essential to operate autonomous vehicles, drones and robotic applications where human performance is impeded the most. A large body of work explores removing weather-induced image degradations with dehazing methods. Most methods rely on single images as input and struggle to generalize from synthetic fully-supervised training approaches or to generate high fidelity results from unpaired real-world datasets. With data as bottleneck and most of today's training data relying on good weather conditions with inclement weather as outlier, we rely on an inverse rendering approach to reconstruct the scene content. We introduce ScatterNeRF, a neural rendering method which adequately renders foggy scenes and decomposes the fog-free background from the participating media-exploiting the multiple views from a short automotive sequence without the need for a large training data corpus. Instead, the rendering approach is optimized on the multi-view scene itself, which can be typically captured by an autonomous vehicle, robot or drone during operation. Specifically, we propose a disentangled representation for the scattering volume and the scene objects, and learn the scene reconstruction with physics-inspired losses. We validate our method by capturing multi-view In-the-Wild data and controlled captures in a large-scale fog chamber.

Seung-Hwan Baek, Felix Heide

Time-of-flight (ToF) sensors provide an imaging modality fueling diverse applications, including LiDAR in autonomous driving, robotics, and augmented reality. Conventional ToF imaging methods estimate the depth by sending pulses of light into a scene and measuring the ToF of the first-arriving photons directly reflected from a scene surface without any temporal delay. As such, all photons following this first response are typically considered as unwanted noise. In this paper, we depart from the principle of using first-arriving photons and propose an all-photon ToF imaging method by incorporating the temporal-polarimetric analysis of first- and late-arriving photons, which possess rich scene information about its geometry and material. To this end, we propose a novel temporal-polarimetric reflectance model, an efficient capture method, and a reconstruction method that exploits the temporal-polarimetric changes of light reflected by the surface and sub-surface reflection. The proposed all-photon polarimetric ToF imaging method allows for acquiring depth, surface normals, and material parameters of a scene by utilizing all photons captured by the system, whereas conventional ToF imaging only obtains coarse depth from the first-arriving photons. We validate our method in simulation and experimentally with a prototype.