Ersin Yumer

Tejas Panambur, Ishan Rajendrakumar Dave, Chongjian Ge et al.

Modern text-to-video (T2V) diffusion models can synthesize visually compelling clips, yet they remain brittle at fine-scale structure: even state-of-the-art generators often produce distorted faces and hands, warped backgrounds, and temporally inconsistent motion. Such severe structural artifacts also appear in very low-quality real-world videos. Classical video restoration and super-resolution (VR/VSR) methods, in contrast, are tuned for synthetic degradations such as blur and downsampling and tend to stabilize these artifacts rather than repair them, while diffusion-prior restorers are usually trained on photometric noise and offer little control over the trade-off between perceptual quality and fidelity. We introduce CreativeVR, a diffusion-prior-guided video restoration framework for AI-generated (AIGC) and real videos with severe structural and temporal artifacts. Our deep-adapter-based method exposes a single precision knob that controls how strongly the model follows the input, smoothly trading off between precise restoration on standard degradations and stronger structure- and motion-corrective behavior on challenging content. Our key novelty is a temporally coherent degradation module used during training, which applies carefully designed transformations that produce realistic structural failures. To evaluate AIGC-artifact restoration, we propose the AIGC54 benchmark with FIQA, semantic and perceptual metrics, and multi-aspect scoring. CreativeVR achieves state-of-the-art results on videos with severe artifacts and performs competitively on standard video restoration benchmarks, while running at practical throughput (about 13 FPS at 720p on a single 80-GB A100). Project page: https://daveishan.github.io/creativevr-webpage/.

Sara Ghazanfari, Wei-An Lin, Haitong Tian et al.

Visually-guided image editing, where edits are conditioned on both visual cues and textual prompts, has emerged as a powerful paradigm for fine-grained, controllable content generation. Although recent generative models have shown remarkable capabilities, existing evaluations remain simple and insufficiently representative of real-world editing challenges. We present SpotEdit, a comprehensive benchmark designed to systematically assess visually-guided image editing methods across diverse diffusion, autoregressive, and hybrid generative models, uncovering substantial performance disparities. To address a critical yet underexplored challenge, our benchmark includes a dedicated component on hallucination, highlighting how leading models, such as GPT-4o, often hallucinate the existence of a visual cue and erroneously perform the editing task. Our code and benchmark are publicly released at https://github.com/SaraGhazanfari/SpotEdit.

Yuwen Xiong, Renjie Liao, Hengshuang Zhao et al.

In this paper, we propose a unified panoptic segmentation network (UPSNet) for tackling the newly proposed panoptic segmentation task. On top of a single backbone residual network, we first design a deformable convolution based semantic segmentation head and a Mask R-CNN style instance segmentation head which solve these two subtasks simultaneously. More importantly, we introduce a parameter-free panoptic head which solves the panoptic segmentation via pixel-wise classification. It first leverages the logits from the previous two heads and then innovatively expands the representation for enabling prediction of an extra unknown class which helps better resolve the conflicts between semantic and instance segmentation. Additionally, it handles the challenge caused by the varying number of instances and permits back propagation to the bottom modules in an end-to-end manner. Extensive experimental results on Cityscapes, COCO and our internal dataset demonstrate that our UPSNet achieves state-of-the-art performance with much faster inference. Code has been made available at: https://github.com/uber-research/UPSNet

Chen Liu, Jimei Yang, Duygu Ceylan et al.

This paper proposes a deep neural network (DNN) for piece-wise planar depthmap reconstruction from a single RGB image. While DNNs have brought remarkable progress to single-image depth prediction, piece-wise planar depthmap reconstruction requires a structured geometry representation, and has been a difficult task to master even for DNNs. The proposed end-to-end DNN learns to directly infer a set of plane parameters and corresponding plane segmentation masks from a single RGB image. We have generated more than 50,000 piece-wise planar depthmaps for training and testing from ScanNet, a large-scale RGBD video database. Our qualitative and quantitative evaluations demonstrate that the proposed approach outperforms baseline methods in terms of both plane segmentation and depth estimation accuracy. To the best of our knowledge, this paper presents the first end-to-end neural architecture for piece-wise planar reconstruction from a single RGB image. Code and data are available at https://github.com/art-programmer/PlaneNet.

Min Bai, Shenlong Wang, Kelvin Wong et al.

In this paper, we introduce a non-parametric memory representation for spatio-temporal segmentation that captures the local space and time around an autonomous vehicle (AV). Our representation has three important properties: (i) it remembers what it has seen in the past, (ii) it reinforces and (iii) forgets its past beliefs based on new evidence. Reinforcing is important as the first time we see an element we might be uncertain, e.g, if the element is heavily occluded or at range. Forgetting is desirable, as otherwise false positives will make the self driving vehicle behave erratically. Our process is informed by 3D reasoning, as occlusion is key to distinguishing between the desire to forget and to remember. We show how our method can be used as an online component to complement static world representations such as HD maps by detecting and remembering changes that should be superimposed on top of this static view due to such events.

Jerry Liu, Wenyuan Zeng, Raquel Urtasun et al.

An intelligent agent operating in the real-world must balance achieving its goal with maintaining the safety and comfort of not only itself, but also other participants within the surrounding scene. This requires jointly reasoning about the behavior of other actors while deciding its own actions as these two processes are inherently intertwined - a vehicle will yield to us if we decide to proceed first at the intersection but will proceed first if we decide to yield. However, this is not captured in most self-driving pipelines, where planning follows prediction. In this paper we propose a novel data-driven, reactive planning objective which allows a self-driving vehicle to jointly reason about its own plans as well as how other actors will react to them. We formulate the problem as an energy-based deep structured model that is learned from observational data and encodes both the planning and prediction problems. Through simulations based on both real-world driving and synthetically generated dense traffic, we demonstrate that our reactive model outperforms a non-reactive variant in successfully completing highly complex maneuvers (lane merges/turns in traffic) faster, without trading off collision rate.

James Tu, Huichen Li, Xinchen Yan et al.

Modern self-driving perception systems have been shown to improve upon processing complementary inputs such as LiDAR with images. In isolation, 2D images have been found to be extremely vulnerable to adversarial attacks. Yet, there have been limited studies on the adversarial robustness of multi-modal models that fuse LiDAR features with image features. Furthermore, existing works do not consider physically realizable perturbations that are consistent across the input modalities. In this paper, we showcase practical susceptibilities of multi-sensor detection by placing an adversarial object on top of a host vehicle. We focus on physically realizable and input-agnostic attacks as they are feasible to execute in practice, and show that a single universal adversary can hide different host vehicles from state-of-the-art multi-modal detectors. Our experiments demonstrate that successful attacks are primarily caused by easily corrupted image features. Furthermore, we find that in modern sensor fusion methods which project image features into 3D, adversarial attacks can exploit the projection process to generate false positives across distant regions in 3D. Towards more robust multi-modal perception systems, we show that adversarial training with feature denoising can boost robustness to such attacks significantly. However, we find that standard adversarial defenses still struggle to prevent false positives which are also caused by inaccurate associations between 3D LiDAR points and 2D pixels.

Ze Yang, Shenlong Wang, Sivabalan Manivasagam et al.

Constructing and animating humans is an important component for building virtual worlds in a wide variety of applications such as virtual reality or robotics testing in simulation. As there are exponentially many variations of humans with different shape, pose and clothing, it is critical to develop methods that can automatically reconstruct and animate humans at scale from real world data. Towards this goal, we represent the pedestrian's shape, pose and skinning weights as neural implicit functions that are directly learned from data. This representation enables us to handle a wide variety of different pedestrian shapes and poses without explicitly fitting a human parametric body model, allowing us to handle a wider range of human geometries and topologies. We demonstrate the effectiveness of our approach on various datasets and show that our reconstructions outperform existing state-of-the-art methods. Furthermore, our re-animation experiments show that we can generate 3D human animations at scale from a single RGB image (and/or an optional LiDAR sweep) as input.

Abbas Sadat, Sean Segal, Sergio Casas et al.

Modern self-driving autonomy systems heavily rely on deep learning. As a consequence, their performance is influenced significantly by the quality and richness of the training data. Data collecting platforms can generate many hours of raw data in a daily basis, however, it is not feasible to label everything. It is thus of key importance to have a mechanism to identify "what to label". Active learning approaches identify examples to label, but their interestingness is tied to a fixed model performing a particular task. These assumptions are not valid in self-driving, where we have to solve a diverse set of tasks (i.e., perception, and motion forecasting) and our models evolve over time frequently. In this paper we introduce a novel approach and propose a new data selection method that exploits a diverse set of criteria that quantize interestingness of traffic scenes. Our experiments on a wide range of tasks and models show that the proposed curation pipeline is able to select datasets that lead to better generalization and higher performance.

Yun Chen, Frieda Rong, Shivam Duggal et al.

Scalable sensor simulation is an important yet challenging open problem for safety-critical domains such as self-driving. Current works in image simulation either fail to be photorealistic or do not model the 3D environment and the dynamic objects within, losing high-level control and physical realism. In this paper, we present GeoSim, a geometry-aware image composition process which synthesizes novel urban driving scenarios by augmenting existing images with dynamic objects extracted from other scenes and rendered at novel poses. Towards this goal, we first build a diverse bank of 3D objects with both realistic geometry and appearance from sensor data. During simulation, we perform a novel geometry-aware simulation-by-composition procedure which 1) proposes plausible and realistic object placements into a given scene, 2) render novel views of dynamic objects from the asset bank, and 3) composes and blends the rendered image segments. The resulting synthetic images are realistic, traffic-aware, and geometrically consistent, allowing our approach to scale to complex use cases. We demonstrate two such important applications: long-range realistic video simulation across multiple camera sensors, and synthetic data generation for data augmentation on downstream segmentation tasks. Please check https://tmux.top/publication/geosim/ for high-resolution video results.

Sean Segal, Eric Kee, Wenjie Luo et al.

In this paper, we tackle the problem of spatio-temporal tagging of self-driving scenes from raw sensor data. Our approach learns a universal embedding for all tags, enabling efficient tagging of many attributes and faster learning of new attributes with limited data. Importantly, the embedding is spatio-temporally aware, allowing the model to naturally output spatio-temporal tag values. Values can then be pooled over arbitrary regions, in order to, for example, compute the pedestrian density in front of the SDV, or determine if a car is blocking another car at a 4-way intersection. We demonstrate the effectiveness of our approach on a new large scale self-driving dataset, SDVScenes, containing 15 attributes relating to vehicle and pedestrian density, the actions of each actor, the speed of each actor, interactions between actors, and the topology of the road map.

Cong Han Lim, Raquel Urtasun, Ersin Yumer

We introduce a new framework for the exact point-wise $\ell_p$ robustness verification problem that exploits the layer-wise geometric structure of deep feed-forward networks with rectified linear activations (ReLU networks). The activation regions of the network partition the input space, and one can verify the $\ell_p$ robustness around a point by checking all the activation regions within the desired radius. The GeoCert algorithm (Jordan et al., NeurIPS 2019) treats this partition as a generic polyhedral complex in order to detect which region to check next. In contrast, our LayerCert framework considers the \emph{nested hyperplane arrangement} structure induced by the layers of the ReLU network and explores regions in a hierarchical manner. We show that, under certain conditions on the algorithm parameters, LayerCert provably reduces the number and size of the convex programs that one needs to solve compared to GeoCert. Furthermore, our LayerCert framework allows the incorporation of lower bounding routines based on convex relaxations to further improve performance. Experimental results demonstrate that LayerCert can significantly reduce both the number of convex programs solved and the running time over the state-of-the-art.

Kibok Lee, Zhuoyuan Chen, Xinchen Yan et al.

We introduce ShapeAdv, a novel framework to study shape-aware adversarial perturbations that reflect the underlying shape variations (e.g., geometric deformations and structural differences) in the 3D point cloud space. We develop shape-aware adversarial 3D point cloud attacks by leveraging the learned latent space of a point cloud auto-encoder where the adversarial noise is applied in the latent space. Specifically, we propose three different variants including an exemplar-based one by guiding the shape deformation with auxiliary data, such that the generated point cloud resembles the shape morphing between objects in the same category. Different from prior works, the resulting adversarial 3D point clouds reflect the shape variations in the 3D point cloud space while still being close to the original one. In addition, experimental evaluations on the ModelNet40 benchmark demonstrate that our adversaries are more difficult to defend with existing point cloud defense methods and exhibit a higher attack transferability across classifiers. Our shape-aware adversarial attacks are orthogonal to existing point cloud based attacks and shed light on the vulnerability of 3D deep neural networks.

Abbas Sadat, Mengye Ren, Andrei Pokrovsky et al.

The motion planners used in self-driving vehicles need to generate trajectories that are safe, comfortable, and obey the traffic rules. This is usually achieved by two modules: behavior planner, which handles high-level decisions and produces a coarse trajectory, and trajectory planner that generates a smooth, feasible trajectory for the duration of the planning horizon. These planners, however, are typically developed separately, and changes in the behavior planner might affect the trajectory planner in unexpected ways. Furthermore, the final trajectory outputted by the trajectory planner might differ significantly from the one generated by the behavior planner, as they do not share the same objective. In this paper, we propose a jointly learnable behavior and trajectory planner. Unlike most existing learnable motion planners that address either only behavior planning, or use an uninterpretable neural network to represent the entire logic from sensors to driving commands, our approach features an interpretable cost function on top of perception, prediction and vehicle dynamics, and a joint learning algorithm that learns a shared cost function employed by our behavior and trajectory components. Experiments on real-world self-driving data demonstrate that jointly learned planner performs significantly better in terms of both similarity to human driving and other safety metrics, compared to baselines that do not adopt joint behavior and trajectory learning.

Mengtian Li, Ersin Yumer, Deva Ramanan

In most practical settings and theoretical analyses, one assumes that a model can be trained until convergence. However, the growing complexity of machine learning datasets and models may violate such assumptions. Indeed, current approaches for hyper-parameter tuning and neural architecture search tend to be limited by practical resource constraints. Therefore, we introduce a formal setting for studying training under the non-asymptotic, resource-constrained regime, i.e., budgeted training. We analyze the following problem: "given a dataset, algorithm, and fixed resource budget, what is the best achievable performance?" We focus on the number of optimization iterations as the representative resource. Under such a setting, we show that it is critical to adjust the learning rate schedule according to the given budget. Among budget-aware learning schedules, we find simple linear decay to be both robust and high-performing. We support our claim through extensive experiments with state-of-the-art models on ImageNet (image classification), Kinetics (video classification), MS COCO (object detection and instance segmentation), and Cityscapes (semantic segmentation). We also analyze our results and find that the key to a good schedule is budgeted convergence, a phenomenon whereby the gradient vanishes at the end of each allowed budget. We also revisit existing approaches for fast convergence and show that budget-aware learning schedules readily outperform such approaches under (the practical but under-explored) budgeted training setting.

Mengtian Li, Zhe Lin, Radomir Mech et al.

Edges, boundaries and contours are important subjects of study in both computer graphics and computer vision. On one hand, they are the 2D elements that convey 3D shapes, on the other hand, they are indicative of occlusion events and thus separation of objects or semantic concepts. In this paper, we aim to generate contour drawings, boundary-like drawings that capture the outline of the visual scene. Prior art often cast this problem as boundary detection. However, the set of visual cues presented in the boundary detection output are different from the ones in contour drawings, and also the artistic style is ignored. We address these issues by collecting a new dataset of contour drawings and proposing a learning-based method that resolves diversity in the annotation and, unlike boundary detectors, can work with imperfect alignment of the annotation and the actual ground truth. Our method surpasses previous methods quantitatively and qualitatively. Surprisingly, when our model fine-tunes on BSDS500, we achieve the state-of-the-art performance in salient boundary detection, suggesting contour drawing might be a scalable alternative to boundary annotation, which at the same time is easier and more interesting for annotators to draw.

Xinchen Yan, Akash Rastogi, Ruben Villegas et al.

Long-term human motion can be represented as a series of motion modes---motion sequences that capture short-term temporal dynamics---with transitions between them. We leverage this structure and present a novel Motion Transformation Variational Auto-Encoders (MT-VAE) for learning motion sequence generation. Our model jointly learns a feature embedding for motion modes (that the motion sequence can be reconstructed from) and a feature transformation that represents the transition of one motion mode to the next motion mode. Our model is able to generate multiple diverse and plausible motion sequences in the future from the same input. We apply our approach to both facial and full body motion, and demonstrate applications like analogy-based motion transfer and video synthesis.

Wei-Sheng Lai, Jia-Bin Huang, Oliver Wang et al.

Applying image processing algorithms independently to each frame of a video often leads to undesired inconsistent results over time. Developing temporally consistent video-based extensions, however, requires domain knowledge for individual tasks and is unable to generalize to other applications. In this paper, we present an efficient end-to-end approach based on deep recurrent network for enforcing temporal consistency in a video. Our method takes the original unprocessed and per-frame processed videos as inputs to produce a temporally consistent video. Consequently, our approach is agnostic to specific image processing algorithms applied on the original video. We train the proposed network by minimizing both short-term and long-term temporal losses as well as the perceptual loss to strike a balance between temporal stability and perceptual similarity with the processed frames. At test time, our model does not require computing optical flow and thus achieves real-time speed even for high-resolution videos. We show that our single model can handle multiple and unseen tasks, including but not limited to artistic style transfer, enhancement, colorization, image-to-image translation and intrinsic image decomposition. Extensive objective evaluation and subject study demonstrate that the proposed approach performs favorably against the state-of-the-art methods on various types of videos.

Minh Vo, Ersin Yumer, Kalyan Sunkavalli et al.

Reliable markerless motion tracking of people participating in a complex group activity from multiple moving cameras is challenging due to frequent occlusions, strong viewpoint and appearance variations, and asynchronous video streams. To solve this problem, reliable association of the same person across distant viewpoints and temporal instances is essential. We present a self-supervised framework to adapt a generic person appearance descriptor to the unlabeled videos by exploiting motion tracking, mutual exclusion constraints, and multi-view geometry. The adapted discriminative descriptor is used in a tracking-by-clustering formulation. We validate the effectiveness of our descriptor learning on WILDTRACK [14] and three new complex social scenes captured by multiple cameras with up to 60 people "in the wild". We report significant improvement in association accuracy (up to 18%) and stable and coherent 3D human skeleton tracking (5 to 10 times) over the baseline. Using the reconstructed 3D skeletons, we cut the input videos into a multi-angle video where the image of a specified person is shown from the best visible front-facing camera. Our algorithm detects inter-human occlusion to determine the camera switching moment while still maintaining the flow of the action well.

Gül Varol, Duygu Ceylan, Bryan Russell et al.

Human shape estimation is an important task for video editing, animation and fashion industry. Predicting 3D human body shape from natural images, however, is highly challenging due to factors such as variation in human bodies, clothing and viewpoint. Prior methods addressing this problem typically attempt to fit parametric body models with certain priors on pose and shape. In this work we argue for an alternative representation and propose BodyNet, a neural network for direct inference of volumetric body shape from a single image. BodyNet is an end-to-end trainable network that benefits from (i) a volumetric 3D loss, (ii) a multi-view re-projection loss, and (iii) intermediate supervision of 2D pose, 2D body part segmentation, and 3D pose. Each of them results in performance improvement as demonstrated by our experiments. To evaluate the method, we fit the SMPL model to our network output and show state-of-the-art results on the SURREAL and Unite the People datasets, outperforming recent approaches. Besides achieving state-of-the-art performance, our method also enables volumetric body-part segmentation.

Chen-Hsuan Lin, Ersin Yumer, Oliver Wang et al.

We address the problem of finding realistic geometric corrections to a foreground object such that it appears natural when composited into a background image. To achieve this, we propose a novel Generative Adversarial Network (GAN) architecture that utilizes Spatial Transformer Networks (STNs) as the generator, which we call Spatial Transformer GANs (ST-GANs). ST-GANs seek image realism by operating in the geometric warp parameter space. In particular, we exploit an iterative STN warping scheme and propose a sequential training strategy that achieves better results compared to naive training of a single generator. One of the key advantages of ST-GAN is its applicability to high-resolution images indirectly since the predicted warp parameters are transferable between reference frames. We demonstrate our approach in two applications: (1) visualizing how indoor furniture (e.g. from product images) might be perceived in a room, (2) hallucinating how accessories like glasses would look when matched with real portraits.

Hsiao-Yu Fish Tung, Hsiao-Wei Tung, Ersin Yumer et al.

Current state-of-the-art solutions for motion capture from a single camera are optimization driven: they optimize the parameters of a 3D human model so that its re-projection matches measurements in the video (e.g. person segmentation, optical flow, keypoint detections etc.). Optimization models are susceptible to local minima. This has been the bottleneck that forced using clean green-screen like backgrounds at capture time, manual initialization, or switching to multiple cameras as input resource. In this work, we propose a learning based motion capture model for single camera input. Instead of optimizing mesh and skeleton parameters directly, our model optimizes neural network weights that predict 3D shape and skeleton configurations given a monocular RGB video. Our model is trained using a combination of strong supervision from synthetic data, and self-supervision from differentiable rendering of (a) skeletal keypoints, (b) dense 3D mesh motion, and (c) human-background segmentation, in an end-to-end framework. Empirically we show our model combines the best of both worlds of supervised learning and test-time optimization: supervised learning initializes the model parameters in the right regime, ensuring good pose and surface initialization at test time, without manual effort. Self-supervision by back-propagating through differentiable rendering allows (unsupervised) adaptation of the model to the test data, and offers much tighter fit than a pretrained fixed model. We show that the proposed model improves with experience and converges to low-error solutions where previous optimization methods fail.

Moos Hueting, Pradyumna Reddy, Vladimir Kim et al.

Discovering 3D arrangements of objects from single indoor images is important given its many applications including interior design, content creation, etc. Although heavily researched in the recent years, existing approaches break down under medium or heavy occlusion as the core object detection module starts failing in absence of directly visible cues. Instead, we take into account holistic contextual 3D information, exploiting the fact that objects in indoor scenes co-occur mostly in typical near-regular configurations. First, we use a neural network trained on real indoor annotated images to extract 2D keypoints, and feed them to a 3D candidate object generation stage. Then, we solve a global selection problem among these 3D candidates using pairwise co-occurrence statistics discovered from a large 3D scene database. We iterate the process allowing for candidates with low keypoint response to be incrementally detected based on the location of the already discovered nearby objects. Focusing on chairs, we demonstrate significant performance improvement over combinations of state-of-the-art methods, especially for scenes with moderately to severely occluded objects.

Chuhang Zou, Ersin Yumer, Jimei Yang et al.

The success of various applications including robotics, digital content creation, and visualization demand a structured and abstract representation of the 3D world from limited sensor data. Inspired by the nature of human perception of 3D shapes as a collection of simple parts, we explore such an abstract shape representation based on primitives. Given a single depth image of an object, we present 3D-PRNN, a generative recurrent neural network that synthesizes multiple plausible shapes composed of a set of primitives. Our generative model encodes symmetry characteristics of common man-made objects, preserves long-range structural coherence, and describes objects of varying complexity with a compact representation. We also propose a method based on Gaussian Fields to generate a large scale dataset of primitive-based shape representations to train our network. We evaluate our approach on a wide range of examples and show that it outperforms nearest-neighbor based shape retrieval methods and is on-par with voxel-based generative models while using a significantly reduced parameter space.

Guilin Liu, Duygu Ceylan, Ersin Yumer et al.

The ability to edit materials of objects in images is desirable by many content creators. However, this is an extremely challenging task as it requires to disentangle intrinsic physical properties of an image. We propose an end-to-end network architecture that replicates the forward image formation process to accomplish this task. Specifically, given a single image, the network first predicts intrinsic properties, i.e. shape, illumination, and material, which are then provided to a rendering layer. This layer performs in-network image synthesis, thereby enabling the network to understand the physics behind the image formation process. The proposed rendering layer is fully differentiable, supports both diffuse and specular materials, and thus can be applicable in a variety of problem settings. We demonstrate a rich set of visually plausible material editing examples and provide an extensive comparative study.

Haibin Huang, Evangelos Kalogerakis, Siddhartha Chaudhuri et al.

We present a new local descriptor for 3D shapes, directly applicable to a wide range of shape analysis problems such as point correspondences, semantic segmentation, affordance prediction, and shape-to-scan matching. The descriptor is produced by a convolutional network that is trained to embed geometrically and semantically similar points close to one another in descriptor space. The network processes surface neighborhoods around points on a shape that are captured at multiple scales by a succession of progressively zoomed out views, taken from carefully selected camera positions. We leverage two extremely large sources of data to train our network. First, since our network processes rendered views in the form of 2D images, we repurpose architectures pre-trained on massive image datasets. Second, we automatically generate a synthetic dense point correspondence dataset by non-rigid alignment of corresponding shape parts in a large collection of segmented 3D models. As a result of these design choices, our network effectively encodes multi-scale local context and fine-grained surface detail. Our network can be trained to produce either category-specific descriptors or more generic descriptors by learning from multiple shape categories. Once trained, at test time, the network extracts local descriptors for shapes without requiring any part segmentation as input. Our method can produce effective local descriptors even for shapes whose category is unknown or different from the ones used while training. We demonstrate through several experiments that our learned local descriptors are more discriminative compared to state of the art alternatives, and are effective in a variety of shape analysis applications.

Xiao Yang, Ersin Yumer, Paul Asente et al.

We present an end-to-end, multimodal, fully convolutional network for extracting semantic structures from document images. We consider document semantic structure extraction as a pixel-wise segmentation task, and propose a unified model that classifies pixels based not only on their visual appearance, as in the traditional page segmentation task, but also on the content of underlying text. Moreover, we propose an efficient synthetic document generation process that we use to generate pretraining data for our network. Once the network is trained on a large set of synthetic documents, we fine-tune the network on unlabeled real documents using a semi-supervised approach. We systematically study the optimum network architecture and show that both our multimodal approach and the synthetic data pretraining significantly boost the performance.

Jun Li, Kai Xu, Siddhartha Chaudhuri et al.

We introduce a novel neural network architecture for encoding and synthesis of 3D shapes, particularly their structures. Our key insight is that 3D shapes are effectively characterized by their hierarchical organization of parts, which reflects fundamental intra-shape relationships such as adjacency and symmetry. We develop a recursive neural net (RvNN) based autoencoder to map a flat, unlabeled, arbitrary part layout to a compact code. The code effectively captures hierarchical structures of man-made 3D objects of varying structural complexities despite being fixed-dimensional: an associated decoder maps a code back to a full hierarchy. The learned bidirectional mapping is further tuned using an adversarial setup to yield a generative model of plausible structures, from which novel structures can be sampled. Finally, our structure synthesis framework is augmented by a second trained module that produces fine-grained part geometry, conditioned on global and local structural context, leading to a full generative pipeline for 3D shapes. We demonstrate that without supervision, our network learns meaningful structural hierarchies adhering to perceptual grouping principles, produces compact codes which enable applications such as shape classification and partial matching, and supports shape synthesis and interpolation with significant variations in topology and geometry.

Zhixin Shu, Ersin Yumer, Sunil Hadap et al.

Traditional face editing methods often require a number of sophisticated and task specific algorithms to be applied one after the other --- a process that is tedious, fragile, and computationally intensive. In this paper, we propose an end-to-end generative adversarial network that infers a face-specific disentangled representation of intrinsic face properties, including shape (i.e. normals), albedo, and lighting, and an alpha matte. We show that this network can be trained on "in-the-wild" images by incorporating an in-network physically-based image formation module and appropriate loss functions. Our disentangling latent representation allows for semantically relevant edits, where one aspect of facial appearance can be manipulated while keeping orthogonal properties fixed, and we demonstrate its use for a number of facial editing applications.

Marc-André Gardner, Kalyan Sunkavalli, Ersin Yumer et al.

We propose an automatic method to infer high dynamic range illumination from a single, limited field-of-view, low dynamic range photograph of an indoor scene. In contrast to previous work that relies on specialized image capture, user input, and/or simple scene models, we train an end-to-end deep neural network that directly regresses a limited field-of-view photo to HDR illumination, without strong assumptions on scene geometry, material properties, or lighting. We show that this can be accomplished in a three step process: 1) we train a robust lighting classifier to automatically annotate the location of light sources in a large dataset of LDR environment maps, 2) we use these annotations to train a deep neural network that predicts the location of lights in a scene from a single limited field-of-view photo, and 3) we fine-tune this network using a small dataset of HDR environment maps to predict light intensities. This allows us to automatically recover high-quality HDR illumination estimates that significantly outperform previous state-of-the-art methods. Consequently, using our illumination estimates for applications like 3D object insertion, we can achieve results that are photo-realistic, which is validated via a perceptual user study.

Eunbyung Park, Jimei Yang, Ersin Yumer et al.

We present a transformation-grounded image generation network for novel 3D view synthesis from a single image. Instead of taking a 'blank slate' approach, we first explicitly infer the parts of the geometry visible both in the input and novel views and then re-cast the remaining synthesis problem as image completion. Specifically, we both predict a flow to move the pixels from the input to the novel view along with a novel visibility map that helps deal with occulsion/disocculsion. Next, conditioned on those intermediate results, we hallucinate (infer) parts of the object invisible in the input image. In addition to the new network structure, training with a combination of adversarial and perceptual loss results in a reduction in common artifacts of novel view synthesis such as distortions and holes, while successfully generating high frequency details and preserving visual aspects of the input image. We evaluate our approach on a wide range of synthetic and real examples. Both qualitative and quantitative results show our method achieves significantly better results compared to existing methods.

Yinda Zhang, Shuran Song, Ersin Yumer et al.

Indoor scene understanding is central to applications such as robot navigation and human companion assistance. Over the last years, data-driven deep neural networks have outperformed many traditional approaches thanks to their representation learning capabilities. One of the bottlenecks in training for better representations is the amount of available per-pixel ground truth data that is required for core scene understanding tasks such as semantic segmentation, normal prediction, and object edge detection. To address this problem, a number of works proposed using synthetic data. However, a systematic study of how such synthetic data is generated is missing. In this work, we introduce a large-scale synthetic dataset with 400K physically-based rendered images from 45K realistic 3D indoor scenes. We study the effects of rendering methods and scene lighting on training for three computer vision tasks: surface normal prediction, semantic segmentation, and object boundary detection. This study provides insights into the best practices for training with synthetic data (more realistic rendering is worth it) and shows that pretraining with our new synthetic dataset can improve results beyond the current state of the art on all three tasks.

Xinchen Yan, Jimei Yang, Ersin Yumer et al.

Understanding the 3D world is a fundamental problem in computer vision. However, learning a good representation of 3D objects is still an open problem due to the high dimensionality of the data and many factors of variation involved. In this work, we investigate the task of single-view 3D object reconstruction from a learning agent's perspective. We formulate the learning process as an interaction between 3D and 2D representations and propose an encoder-decoder network with a novel projection loss defined by the perspective transformation. More importantly, the projection loss enables the unsupervised learning using 2D observation without explicit 3D supervision. We demonstrate the ability of the model in generating 3D volume from a single 2D image with three sets of experiments: (1) learning from single-class objects; (2) learning from multi-class objects and (3) testing on novel object classes. Results show superior performance and better generalization ability for 3D object reconstruction when the projection loss is involved.