Search

Long-horizon contact-rich manipulation has long been a challenging problem, as it requires reasoning over both discrete contact modes and continuous object motion. We introduce Implicit Contact Diffuser (ICD), a diffusion-based model that generates a sequence of neural descriptors that specify a series of contact relationships between the object and the environment. This sequence is then used as guidance for an MPC method to accomplish a given task. The key advantage of this approach is that the latent descriptors provide more task-relevant guidance to MPC, helping to avoid local minima for contact-rich manipulation tasks. Our experiments demonstrate that ICD outperforms baselines on complex, long-horizon, contact-rich manipulation tasks, such as cable routing and notebook folding. Additionally, our experiments also indicate that \methodshort can generalize a target contact relationship to a different environment. More visualizations can be found on our website $\href{https://implicit-contact-diffuser.github.io/}{https://implicit-contact-diffuser.github.io}$, Comment: In submussion

We present SF3D, a novel method for rapid and high-quality textured object mesh reconstruction from a single image in just 0.5 seconds. Unlike most existing approaches, SF3D is explicitly trained for mesh generation, incorporating a fast UV unwrapping technique that enables swift texture generation rather than relying on vertex colors. The method also learns to predict material parameters and normal maps to enhance the visual quality of the reconstructed 3D meshes. Furthermore, SF3D integrates a delighting step to effectively remove low-frequency illumination effects, ensuring that the reconstructed meshes can be easily used in novel illumination conditions. Experiments demonstrate the superior performance of SF3D over the existing techniques. Project page: https://stable-fast-3d.github.io

In this paper, we study efficient multi-beam training design for near-field communications to reduce the beam training overhead of conventional single-beam training methods. In particular, the array-division based multi-beam training method, which is widely used in far-field communications, cannot be directly applied to the near-field scenario, since different sub-arrays may observe different user angles and there exist coverage holes in the angular domain. To address these issues, we first devise a new near-field multi-beam codebook by sparsely activating a portion of antennas to form a sparse linear array (SLA), hence generating multiple beams simultaneously by effective exploiting the near-field grating-lobs. Next, a two-stage near-field beam training method is proposed, for which several candidate user locations are identified firstly based on multi-beam sweeping over time, followed by the second stage to further determine the true user location with a small number of single-beam sweeping. Finally, numerical results show that our proposed multi-beam training method significantly reduces the beam training overhead of conventional single-beam training methods, yet achieving comparable rate performance in data transmission., Comment: In this paper, a novel near-field multi-beam training scheme is proposed by sparsely activating a portion of antennas to form a sparse linear array

We present PointInfinity, an efficient family of point cloud diffusion models. Our core idea is to use a transformer-based architecture with a fixed-size, resolution-invariant latent representation. This enables efficient training with low-resolution point clouds, while allowing high-resolution point clouds to be generated during inference. More importantly, we show that scaling the test-time resolution beyond the training resolution improves the fidelity of generated point clouds and surfaces. We analyze this phenomenon and draw a link to classifier-free guidance commonly used in diffusion models, demonstrating that both allow trading off fidelity and variability during inference. Experiments on CO3D show that PointInfinity can efficiently generate high-resolution point clouds (up to 131k points, 31 times more than Point-E) with state-of-the-art quality., Comment: Accepted to CVPR 2024, project website at https://zixuanh.com/projects/pointinfinity

Manipulation of articulated and deformable objects can be difficult due to their compliant and under-actuated nature. Unexpected disturbances can cause the object to deviate from a predicted state, making it necessary to use Model-Predictive Control (MPC) methods to plan motion. However, these methods need a short planning horizon to be practical. Thus, MPC is ill-suited for long-horizon manipulation tasks due to local minima. In this paper, we present a diffusion-based method that guides an MPC method to accomplish long-horizon manipulation tasks by dynamically specifying sequences of subgoals for the MPC to follow. Our method, called Subgoal Diffuser, generates subgoals in a coarse-to-fine manner, producing sparse subgoals when the task is easily accomplished by MPC and more dense subgoals when the MPC method needs more guidance. The density of subgoals is determined dynamically based on a learned estimate of reachability, and subgoals are distributed to focus on challenging parts of the task. We evaluate our method on two robot manipulation tasks and find it improves the planning performance of an MPC method, and also outperforms prior diffusion-based methods., Comment: ICRA 2024

This technical report introduces TripoSR, a 3D reconstruction model leveraging transformer architecture for fast feed-forward 3D generation, producing 3D mesh from a single image in under 0.5 seconds. Building upon the LRM network architecture, TripoSR integrates substantial improvements in data processing, model design, and training techniques. Evaluations on public datasets show that TripoSR exhibits superior performance, both quantitatively and qualitatively, compared to other open-source alternatives. Released under the MIT license, TripoSR is intended to empower researchers, developers, and creatives with the latest advancements in 3D generative AI., Comment: Model: https://huggingface.co/stabilityai/TripoSR Code: https://github.com/VAST-AI-Research/TripoSR Demo: https://huggingface.co/spaces/stabilityai/TripoSR

The prominent large language models (LLMs) of today differ from past language models not only in size, but also in the fact that they are trained on a combination of natural language and formal language (code). As a medium between humans and computers, code translates high-level goals into executable steps, featuring standard syntax, logical consistency, abstraction, and modularity. In this survey, we present an overview of the various benefits of integrating code into LLMs' training data. Specifically, beyond enhancing LLMs in code generation, we observe that these unique properties of code help (i) unlock the reasoning ability of LLMs, enabling their applications to a range of more complex natural language tasks; (ii) steer LLMs to produce structured and precise intermediate steps, which can then be connected to external execution ends through function calls; and (iii) take advantage of code compilation and execution environment, which also provides diverse feedback for model improvement. In addition, we trace how these profound capabilities of LLMs, brought by code, have led to their emergence as intelligent agents (IAs) in situations where the ability to understand instructions, decompose goals, plan and execute actions, and refine from feedback are crucial to their success on downstream tasks. Finally, we present several key challenges and future directions of empowering LLMs with code.

We study the problem of single-image zero-shot 3D shape reconstruction. Recent works learn zero-shot shape reconstruction through generative modeling of 3D assets, but these models are computationally expensive at train and inference time. In contrast, the traditional approach to this problem is regression-based, where deterministic models are trained to directly regress the object shape. Such regression methods possess much higher computational efficiency than generative methods. This raises a natural question: is generative modeling necessary for high performance, or conversely, are regression-based approaches still competitive? To answer this, we design a strong regression-based model, called ZeroShape, based on the converging findings in this field and a novel insight. We also curate a large real-world evaluation benchmark, with objects from three different real-world 3D datasets. This evaluation benchmark is more diverse and an order of magnitude larger than what prior works use to quantitatively evaluate their models, aiming at reducing the evaluation variance in our field. We show that ZeroShape not only achieves superior performance over state-of-the-art methods, but also demonstrates significantly higher computational and data efficiency., Comment: Project page: https://zixuanh.com/projects/zeroshape.html

This paper introduces Low-shot Object Learning with Mutual Exclusivity Bias (LSME), the first computational framing of mutual exclusivity bias, a phenomenon commonly observed in infants during word learning. We provide a novel dataset, comprehensive baselines, and a state-of-the-art method to enable the ML community to tackle this challenging learning task. The goal of LSME is to analyze an RGB image of a scene containing multiple objects and correctly associate a previously-unknown object instance with a provided category label. This association is then used to perform low-shot learning to test category generalization. We provide a data generation pipeline for the LSME problem and conduct a thorough analysis of the factors that contribute to its difficulty. Additionally, we evaluate the performance of multiple baselines, including state-of-the-art foundation models. Finally, we present a baseline approach that outperforms state-of-the-art models in terms of low-shot accuracy., Comment: Accepted at NeurIPS 2023, Datasets and Benchmarks Track. Project website https://ngailapdi.github.io/projects/lsme/

Intelligent reflecting/refracting surface (IRS) is envisioned as a promising technology to reconfigure wireless propagation environment for enhancing the communication performance, by smartly controlling the signal reflection/refraction with a large number of tunable passive elements. In particular, the application of IRS in high-mobility scenarios can convert wireless channels from fast fading to slow fading, thus achieving more reliable communications. In this paper, we first provide an overview of the new applications and opportunities of IRS in high-mobility communications. Next, we present two practical strategies for deploying IRS to aid high-mobility communications, namely, roadside IRS versus vehicle-side IRS, and compare their different channel characteristics, handover requirements, and deployment costs. Then, the main issues in designing IRS-aided high-mobility communications, including node discovery, mode switching, beam alignment/tracking, handover, and multiuser scheduling are discussed for both IRS deployment strategies. Moreover, numerical results are presented to demonstrate the potential performance gains of IRSs in vehicular communications. Finally, new research directions are pointed out for future work., Comment: Accepted by IEEE Communications Magazine

Captions are crucial for understanding scientific visualizations and documents. Existing captioning methods for scientific figures rely on figure-caption pairs extracted from documents for training, many of which fall short with respect to metrics like helpfulness, explainability, and visual-descriptiveness [15] leading to generated captions being misaligned with reader preferences. To enable the generation of high-quality figure captions, we introduce FigCaps-HF a new framework for figure-caption generation that can incorporate domain expert feedback in generating captions optimized for reader preferences. Our framework comprises of 1) an automatic method for evaluating quality of figure-caption pairs, 2) a novel reinforcement learning with human feedback (RLHF) method to optimize a generative figure-to-caption model for reader preferences. We demonstrate the effectiveness of our simple learning framework by improving performance over standard fine-tuning across different types of models. In particular, when using BLIP as the base model, our RLHF framework achieves a mean gain of 35.7%, 16.9%, and 9% in ROUGE, BLEU, and Meteor, respectively. Finally, we release a large-scale benchmark dataset with human feedback on figure-caption pairs to enable further evaluation and development of RLHF techniques for this problem., Comment: 19 pages, 4 figures. Benchmark Documentation: https://figcapshf.github.io/

We present ShapeClipper, a novel method that reconstructs 3D object shapes from real-world single-view RGB images. Instead of relying on laborious 3D, multi-view or camera pose annotation, ShapeClipper learns shape reconstruction from a set of single-view segmented images. The key idea is to facilitate shape learning via CLIP-based shape consistency, where we encourage objects with similar CLIP encodings to share similar shapes. We also leverage off-the-shelf normals as an additional geometric constraint so the model can learn better bottom-up reasoning of detailed surface geometry. These two novel consistency constraints, when used to regularize our model, improve its ability to learn both global shape structure and local geometric details. We evaluate our method over three challenging real-world datasets, Pix3D, Pascal3D+, and OpenImages, where we achieve superior performance over state-of-the-art methods., Comment: Accepted to CVPR 2023, project website at https://zixuanh.com/projects/shapeclipper.html

State estimation is one of the greatest challenges for cloth manipulation due to cloth's high dimensionality and self-occlusion. Prior works propose to identify the full state of crumpled clothes by training a mesh reconstruction model in simulation. However, such models are prone to suffer from a sim-to-real gap due to differences between cloth simulation and the real world. In this work, we propose a self-supervised method to finetune a mesh reconstruction model in the real world. Since the full mesh of crumpled cloth is difficult to obtain in the real world, we design a special data collection scheme and an action-conditioned model-based cloth tracking method to generate pseudo-labels for self-supervised learning. By finetuning the pretrained mesh reconstruction model on this pseudo-labeled dataset, we show that we can improve the quality of the reconstructed mesh without requiring human annotations, and improve the performance of downstream manipulation task.

A hallmark of the deep learning era for computer vision is the successful use of large-scale labeled datasets to train feature representations for tasks ranging from object recognition and semantic segmentation to optical flow estimation and novel view synthesis of 3D scenes. In this work, we aim to learn dense discriminative object representations for low-shot category recognition without requiring any category labels. To this end, we propose Deep Object Patch Encodings (DOPE), which can be trained from multiple views of object instances without any category or semantic object part labels. To train DOPE, we assume access to sparse depths, foreground masks and known cameras, to obtain pixel-level correspondences between views of an object, and use this to formulate a self-supervised learning task to learn discriminative object patches. We find that DOPE can directly be used for low-shot classification of novel categories using local-part matching, and is competitive with and outperforms supervised and self-supervised learning baselines. Code and data available at https://github.com/rehg-lab/dope_selfsup., Comment: Accepted at NeurIPS 2022. Code and data available at https://github.com/rehg-lab/dope_selfsup

Intelligent reflecting surface (IRS) has emerged as a promising technique to control wireless propagation environment for enhancing the communication performance cost-effectively. However, the rapidly time-varying channel in high-mobility communication scenarios such as vehicular communication renders it challenging to obtain the instantaneous channel state information (CSI) efficiently for IRS with a large number of reflecting elements. In this paper, we propose a new roadside IRS-aided vehicular communication system to tackle this challenge. Specifically, by exploiting the symmetrical deployment of IRSs with inter-laced equal intervals on both sides of the road and the cooperation among nearby IRS controllers, we propose a new two-stage channel estimation scheme with off-line and online training, respectively, to obtain the static/time-varying CSI required by the proposed low-complexity passive beamforming scheme efficiently. The proposed IRS beamforming and online channel estimation designs leverage the existing uplink pilots in wireless networks and do not require any change of the existing transmission protocol. Moreover, they can be implemented by each of IRS controllers independently, without the need of any real-time feedback from the user's serving BS. Simulation results show that the proposed designs can efficiently achieve the high IRS passive beamforming gain and thus significantly enhance the achievable communication throughput for high-speed vehicular communications.

Self-occlusion is challenging for cloth manipulation, as it makes it difficult to estimate the full state of the cloth. Ideally, a robot trying to unfold a crumpled or folded cloth should be able to reason about the cloth's occluded regions. We leverage recent advances in pose estimation for cloth to build a system that uses explicit occlusion reasoning to unfold a crumpled cloth. Specifically, we first learn a model to reconstruct the mesh of the cloth. However, the model will likely have errors due to the complexities of the cloth configurations and due to ambiguities from occlusions. Our main insight is that we can further refine the predicted reconstruction by performing test-time finetuning with self-supervised losses. The obtained reconstructed mesh allows us to use a mesh-based dynamics model for planning while reasoning about occlusions. We evaluate our system both on cloth flattening as well as on cloth canonicalization, in which the objective is to manipulate the cloth into a canonical pose. Our experiments show that our method significantly outperforms prior methods that do not explicitly account for occlusions or perform test-time optimization. Videos and visualizations can be found on our $\href{https://sites.google.com/view/occlusion-reason/home}{\text{project website}}.$, Comment: RSS 2022, $\href{https://sites.google.com/view/occlusion-reason/home}{\text{project website}}$

Skeleton extraction is a task focused on providing a simple representation of an object by extracting the skeleton from the given binary or RGB image. In recent years many attractive works in skeleton extraction have been made. But as far as we know, there is little research on how to utilize the context information in the binary shape of objects. In this paper, we propose an attention-based model called Context Attention Network (CANet), which integrates the context extraction module in a UNet architecture and can effectively improve the ability of network to extract the skeleton pixels. Meanwhile, we also use some novel techniques including distance transform, weight focal loss to achieve good results on the given dataset. Finally, without model ensemble and with only 80% of the training images, our method achieves 0.822 F1 score during the development phase and 0.8507 F1 score during the final phase of the Pixel SkelNetOn Competition, ranking 1st place on the leaderboard., Comment: Accepted at the Deep Learning for Geometric Computing (DLGC) workshop at CVPR 2022

We present a novel 3D shape reconstruction method which learns to predict an implicit 3D shape representation from a single RGB image. Our approach uses a set of single-view images of multiple object categories without viewpoint annotation, forcing the model to learn across multiple object categories without 3D supervision. To facilitate learning with such minimal supervision, we use category labels to guide shape learning with a novel categorical metric learning approach. We also utilize adversarial and viewpoint regularization techniques to further disentangle the effects of viewpoint and shape. We obtain the first results for large-scale (more than 50 categories) single-viewpoint shape prediction using a single model without any 3D cues. We are also the first to examine and quantify the benefit of class information in single-view supervised 3D shape reconstruction. Our method achieves superior performance over state-of-the-art methods on ShapeNet-13, ShapeNet-55 and Pascal3D+., Comment: Project page: https://zixuanh.com/multiclass3D

Global human motion forecasting is important in many fields, which is the combination of global human trajectory prediction and local human pose prediction. Visual and social information are often used to boost model performance, however, they may consume too much computational resource. In this paper, we establish a simple but effective baseline for single human motion forecasting without visual and social information, equipped with useful training tricks. Our method "futuremotion_ICCV21" outperforms existing methods by a large margin on SoMoF benchmark. We hope our work provide new ideas for future research., Comment: ICCV SoMoF Workshop, 2021

Intelligent reflecting/refracting surface (IRS) has recently emerged as a promising solution to reconfigure wireless propagation environment for enhancing the communication performance. In this paper, we study a new IRS-aided high-mobility communication system by employing the intelligent refracting surface with a high-speed vehicle to aid its passenger's communication with a remote base station (BS). Due to the environment's random scattering and vehicle's high mobility, a rapidly time-varying channel is typically resulted between the static BS and fast-moving IRS/user, which renders the channel estimation for IRS with a large number of elements more challenging. In order to reap the high IRS passive beamforming gain with low channel training overhead, we propose a new and efficient transmission protocol to achieve both IRS channel estimation and refraction optimization for data transmission. Specifically, by exploiting the quasi-static channel between the IRS and user both moving at the same high speed as well as the line-of-sight (LoS) dominant channel between the BS and IRS, the user first estimates the LoS component of the cascaded BS-IRS-user channel, based on which IRS passive refraction is designed to maximize the corresponding IRS-refracted channel gain. Then, the user estimates the resultant IRS-refracted channel as well as the non-IRS-refracted channel for setting an additional common phase shift at all IRS refracting elements so as to align these two channels for maximizing the overall channel gain for data transmission. Simulation results show significant performance improvement of the proposed design as compared to various benchmark schemes. The proposed on-vehicle IRS system is further compared with a baseline scheme of deploying fixed intelligent reflecting surfaces on the roadside to assist high-speed vehicular communications, which achieves significant rate improvement., Comment: Conference version (arXiv:2011.03147). Accepted by TWC

Robotic manipulation of cloth remains challenging for robotics due to the complex dynamics of the cloth, lack of a low-dimensional state representation, and self-occlusions. In contrast to previous model-based approaches that learn a pixel-based dynamics model or a compressed latent vector dynamics, we propose to learn a particle-based dynamics model from a partial point cloud observation. To overcome the challenges of partial observability, we infer which visible points are connected on the underlying cloth mesh. We then learn a dynamics model over this visible connectivity graph. Compared to previous learning-based approaches, our model poses strong inductive bias with its particle based representation for learning the underlying cloth physics; it is invariant to visual features; and the predictions can be more easily visualized. We show that our method greatly outperforms previous state-of-the-art model-based and model-free reinforcement learning methods in simulation. Furthermore, we demonstrate zero-shot sim-to-real transfer where we deploy the model trained in simulation on a Franka arm and show that the model can successfully smooth different types of cloth from crumpled configurations. Videos can be found on our project website., Comment: Published at CoRL 2021. Project website: https://sites.google.com/view/vcd-cloth

Continual learning has been extensively studied for classification tasks with methods developed to primarily avoid catastrophic forgetting, a phenomenon where earlier learned concepts are forgotten at the expense of more recent samples. In this work, we present a set of continual 3D object shape reconstruction tasks, including complete 3D shape reconstruction from different input modalities, as well as visible surface (2.5D) reconstruction which, surprisingly demonstrate positive knowledge (backward and forward) transfer when training with solely standard SGD and without additional heuristics. We provide evidence that continuously updated representation learning of single-view 3D shape reconstruction improves the performance on learned and novel categories over time. We provide a novel analysis of knowledge transfer ability by looking at the output distribution shift across sequential learning tasks. Finally, we show that the robustness of these tasks leads to the potential of having a proxy representation learning task for continual classification. The codebase, dataset and pre-trained models released with this article can be found at https://github.com/rehg-lab/CLRec, Comment: Accepted to 3DV 2022

In this paper, we study a new intelligent refracting surface (IRS)-assisted high-mobility communication with the IRS deployed in a high-speed moving vehicle to assist its passenger's communication with a static base station (BS) on the roadside. The vehicle's high Doppler frequency results in a fast fading channel between the BS and the passenger/user, which renders channel estimation for the IRS with a large number of refracting elements a more challenging task as compared to the conventional case with low-mobility users only. In order to mitigate the Doppler effect and reap the full IRS passive beamforming gain with low training overhead, we propose a new and efficient transmission protocol to execute channel estimation and IRS refraction design for data transmission. Specifically, by exploiting the quasi-static channel between the IRS and user both moving at the same high speed, we first estimate the cascaded BS-IRS-user channel with the Doppler effect compensated. Then, we estimate the instantaneous BS-user fast fading channel (without IRS refraction) and tune the IRS refraction over time accordingly to align the cascaded channel with the BS-user direct channel, thus maximizing the IRS's passive beamforming gain as well as converting their combined channel from fast to slow fading. Simulation results show the effectiveness of the proposed channel estimation scheme and passive beamforming design as compared to various benchmark schemes., Comment: Accepted by ICC 2021

We present an interpretable deep model for fine-grained visual recognition. At the core of our method lies the integration of region-based part discovery and attribution within a deep neural network. Our model is trained using image-level object labels, and provides an interpretation of its results via the segmentation of object parts and the identification of their contributions towards classification. To facilitate the learning of object parts without direct supervision, we explore a simple prior of the occurrence of object parts. We demonstrate that this prior, when combined with our region-based part discovery and attribution, leads to an interpretable model that remains highly accurate. Our model is evaluated on major fine-grained recognition datasets, including CUB-200, CelebA and iNaturalist. Our results compare favorably to state-of-the-art methods on classification tasks, and our method outperforms previous approaches on the localization of object parts., Comment: Accepted to CVPR 2020 (Oral)

Recent neural style transfer frameworks have obtained astonishing visual quality and flexibility in Single-style Transfer (SST), but little attention has been paid to Multi-style Transfer (MST) which refers to simultaneously transferring multiple styles to the same image. Compared to SST, MST has the potential to create more diverse and visually pleasing stylization results. In this paper, we propose the first MST framework to automatically incorporate multiple styles into one result based on regional semantics. We first improve the existing SST backbone network by introducing a novel multi-level feature fusion module and a patch attention module to achieve better semantic correspondences and preserve richer style details. For MST, we designed a conceptually simple yet effective region-based style fusion module to insert into the backbone. It assigns corresponding styles to content regions based on semantic matching, and then seamlessly combines multiple styles together. Comprehensive evaluations demonstrate that our framework outperforms existing works of SST and MST., Comment: Pacific Graphics 2019