论文检索

Karl Holmquist, Bastian Wandt

Traditionally, monocular 3D human pose estimation employs a machine learning model to predict the most likely 3D pose for a given input image. However, a single image can be highly ambiguous and induces multiple plausible solutions for the 2D-3D lifting step, which results in overly confident 3D pose predictors. To this end, we propose DiffPose, a conditional diffusion model that predicts multiple hypotheses for a given input image. Compared to similar approaches, our diffusion model is straightforward and avoids intensive hyperparameter tuning, complex network structures, mode collapse, and unstable training. Moreover, we tackle the problem of over-simplification of the intermediate representation of the common two-step approaches which first estimate a distribution of 2D joint locations via joint-wise heatmaps and consecutively use their maximum argument for the 3D pose estimation step. Since such a simplification of the heatmaps removes valid information about possibly correct, though labeled unlikely, joint locations, we propose to represent the heatmaps as a set of 2D joint candidate samples. To extract information about the original distribution from these samples, we introduce our embedding transformer which conditions the diffusion model. Experimentally, we show that DiffPose improves upon the state of the art for multi-hypothesis pose estimation by 3-5% for simple poses and outperforms it by a large margin for highly ambiguous poses.

Chengkai Hou, Jieyu Zhang, Tianyi Zhou

Data augmentation (DA) is widely used to improve the generalization of neural networks by enforcing the invariances and symmetries to pre-defined transformations applied to input data. However, a fixed augmentation policy may have different effects on each sample in different training stages but existing approaches cannot adjust the policy to be adaptive to each sample and the training model. In this paper, we propose "Model-Adaptive Data Augmentation (MADAug)" that jointly trains an augmentation policy network to teach the model "when to learn what". Unlike previous work, MADAug selects augmentation operators for each input image by a model-adaptive policy varying between training stages, producing a data augmentation curriculum optimized for better generalization. In MADAug, we train the policy through a bi-level optimization scheme, which aims to minimize a validation set loss of a model trained using the policy-produced data augmentations. We conduct an extensive evaluation of MADAug on multiple image classification tasks and network architectures with thorough comparisons to existing DA approaches. MADAug outperforms or is on par with other baselines and exhibits better fairness: it brings improvement to all classes and more to the difficult ones. Moreover, MADAug learned policy shows better performance when transferred to fine-grained datasets. In addition, the auto-optimized policy in MADAug gradually introduces increasing perturbations and naturally forms an easy-to-hard curriculum.

Hung Tran, Vuong Le, Svetha Venkatesh, Truyen Tran

Humans are highly adaptable, swiftly switching between different modes to progressively handle different tasks, situations and contexts. In Human-object interaction (HOI) activities, these modes can be attributed to two mechanisms: (1) the large-scale consistent plan for the whole activity and (2) the small-scale children interactive actions that start and end along the timeline. While neuroscience and cognitive science have confirmed this multi-mechanism nature of human behavior, machine modeling approaches for human motion are trailing behind. While attempted to use gradually morphing structures (e.g., graph attention networks) to model the dynamic HOI patterns, they miss the expeditious and discrete mode-switching nature of the human motion. To bridge that gap, this work proposes to model two concurrent mechanisms that jointly control human motion: the Persistent process that runs continually on the global scale, and the Transient sub-processes that operate intermittently on the local context of the human while interacting with objects. These two mechanisms form an interactive Persistent-Transient Duality that synergistically governs the activity sequences. We model this conceptual duality by a parent-child neural network of Persistent and Transient channels with a dedicated neural module for dynamic mechanism switching. The framework is trialed on HOI motion forecasting. On two rich datasets and a wide variety of settings, the model consistently delivers superior performances, proving its suitability for the challenge.

Qingyan Meng, Mingqing Xiao, Shen Yan, Yisen Wang, Zhouchen Lin, Zhi-Quan Luo

Spiking Neural Networks (SNNs) are promising energy-efficient models for neuromorphic computing. For training the non-differentiable SNN models, the backpropagation through time (BPTT) with surrogate gradients (SG) method has achieved high performance. However, this method suffers from considerable memory cost and training time during training. In this paper, we propose the Spatial Learning Through Time (SLTT) method that can achieve high performance while greatly improving training efficiency compared with BPTT. First, we show that the backpropagation of SNNs through the temporal domain contributes just a little to the final calculated gradients. Thus, we propose to ignore the unimportant routes in the computational graph during backpropagation. The proposed method reduces the number of scalar multiplications and achieves a small memory occupation that is independent of the total time steps. Furthermore, we propose a variant of SLTT, called SLTT-K, that allows backpropagation only at K time steps, then the required number of scalar multiplications is further reduced and is independent of the total time steps. Experiments on both static and neuromorphic datasets demonstrate superior training efficiency and performance of our SLTT. In particular, our method achieves state-of-the-art accuracy on ImageNet, while the memory cost and training time are reduced by more than 70% and 50%, respectively, compared with BPTT.

Yuan Liu, Songyang Zhang, Jiacheng Chen, Zhaohui Yu, Kai Chen, Dahua Lin

There has been significant progress in Masked Image Modeling (MIM). Existing MIM methods can be broadly categorized into two groups based on the reconstruction target: pixel-based and tokenizer-based approaches. The former offers a simpler pipeline and lower computational cost, but it is known to be biased toward high-frequency details. In this paper, we provide a set of empirical studies to confirm this limitation of pixel-based MIM and propose a new method that explicitly utilizes low-level features from shallow layers to aid pixel reconstruction. By incorporating this design into our base method, MAE, we reduce the wasted modeling capability of pixel-based MIM, improving its convergence and achieving non-trivial improvements across various downstream tasks. To the best of our knowledge, we are the first to systematically investigate multi-level feature fusion for isotropic architectures like the standard Vision Transformer (ViT). Notably, when applied to a smaller model (e.g., ViT-S), our method yields significant performance gains, such as 1.2% on fine-tuning, 2.8% on linear probing, and 2.6% on semantic segmentation.

Shuiwang Li, Yangxiang Yang, Dan Zeng, Xucheng Wang

While discriminative correlation filters (DCF)-based trackers prevail in UAV tracking for their favorable efficiency, lightweight convolutional neural network (CNN)-based trackers using filter pruning have also demonstrated remarkable efficiency and precision. However, the use of pure vision transformer models (ViTs) for UAV tracking remains unexplored, which is a surprising finding given that ViTs have been shown to produce better performance and greater efficiency than CNNs in image classification. In this paper, we propose an efficient ViT-based tracking framework, Aba-ViTrack, for UAV tracking. In our framework, feature learning and template-search coupling are integrated into an efficient one-stream ViT to avoid an extra heavy relation modeling module. The proposed Aba-ViT exploits an adaptive and background-aware token computation method to reduce inference time. This approach adaptively discards tokens based on learned halting probabilities, which a priori are higher for background tokens than target ones. Extensive experiments on six UAV tracking benchmarks demonstrate that the proposed Aba-ViTrack achieves state-of-the-art performance in UAV tracking. Code is available at https://github.com/xyyang317/Aba-ViTrack.

Xiaoyu Huang, Dhruv Batra, Akshara Rai, Andrew Szot

We present Skill Transformer, an approach for solving long-horizon robotic tasks by combining conditional sequence modeling and skill modularity. Conditioned on egocentric and proprioceptive observations of a robot, Skill Transformer is trained end-to-end to predict both a high-level skill (e.g., navigation, picking, placing), and a whole-body low-level action (e.g., base and arm motion), using a transformer architecture and demonstration trajectories that solve the full task. It retains the composability and modularity of the overall task through a skill predictor module while reasoning about low-level actions and avoiding hand-off errors, common in modular approaches. We test Skill Transformer on an embodied rearrangement benchmark and find it performs robust task planning and low-level control in new scenarios, achieving a 2.5x higher success rate than baselines in hard rearrangement problems.

Kecheng Zheng, Wei Wu, Ruili Feng, Kai Zhu, Jiawei Liu, Deli Zhao, Zheng-Jun Zha, Wei Chen, Yujun Shen

Prompt tuning and adapter tuning have shown great potential in transferring pre-trained vision-language models (VLMs) to various downstream tasks. In this work, we design a new type of tuning method, termed as regularized mask tuning, which masks the network parameters through a learnable selection. Inspired by neural pathways, we argue that the knowledge required by a downstream task already exists in the pre-trained weights but just gets concealed in the upstream pre-training stage. To bring the useful knowledge back into light, we first identify a set of parameters that are important to a given downstream task, then attach a binary mask to each parameter, and finally optimize these masks on the downstream data with the parameters frozen. When updating the mask, we introduce a novel gradient dropout strategy to regularize the parameter selection, in order to prevent the model from forgetting old knowledge and overfitting the downstream data. Experimental results on 11 datasets demonstrate the consistent superiority of our method over previous alternatives. It is noteworthy that we manage to deliver 18.73% performance improvement compared to the zero-shot CLIP via masking an average of only 2.56% parameters. Furthermore, our method is synergistic with most existing parameter-efficient tuning methods and can boost the performance on top of them. Code will be made publicly available.

Alexandre Kirchmeyer, Jia Deng

In computer vision, 2D convolution is arguably the most important operation performed by a ConvNet. Unsurprisingly, it has been the focus of intense software and hardware optimization and enjoys highly efficient implementations. In this work, we ask an intriguing question: can we make a ConvNet work without 2D convolutions? Surprisingly, we find that the answer is yes --- we show that a ConvNet consisting entirely of 1D convolutions can do just as well as 2D on ImageNet classification. Specifically, we find that one key ingredient to a high-performing 1D ConvNet is oriented 1D kernels: 1D kernels that are oriented not just horizontally or vertically, but also at other angles. Our experiments show that oriented 1D convolutions can not only replace 2D convolutions but also augment existing architectures with large kernels, leading to improved accuracy with minimal FLOPs increase. A key contribution of this work is a highly-optimized custom CUDA implementation of oriented 1D kernels, specialized to the depthwise convolution setting. Our benchmarks demonstrate that our custom CUDA implementation almost perfectly realizes the theoretical advantage of 1D convolution: it is faster than a native horizontal convolution for any arbitrary angle. Code is available at https://github.com/princeton-vl/Oriented1D.

Arjun Mani, Ishaan Preetam Chandratreya, Elliot Creager, Carl Vondrick, Richard Zemel

Modeling the mechanics of fluid in complex scenes is vital to applications in design, graphics, and robotics. Learning-based methods provide fast and differentiable fluid simulators, however most prior work is unable to accurately model how fluids interact with genuinely novel surfaces not seen during training. We introduce SurfsUP, a framework that represents objects implicitly using signed distance functions (SDFs), rather than an explicit representation of meshes or particles. This continuous representation of geometry enables more accurate simulation of fluid-object interactions over long time periods while simultaneously making computation more efficient. Moreover, SurfsUP trained on simple shape primitives generalizes considerably out-of-distribution, even to complex real-world scenes and objects. Finally, we show we can invert our model to design simple objects to manipulate fluid flow.

Zijian Wang, Yadan Luo, Liang Zheng, Zi Huang, Mahsa Baktashmotlagh

This paper focuses on model transferability estimation, i.e., assessing the performance of pre-trained models on a downstream task without performing fine-tuning. Motivated by the neural collapse (NC) that reveals the feature geometry at the terminal stage of training, our method considers the model transferability as how far the target activations obtained by pre-trained models are from their hypothetical state in the terminal phase of the fine-tuned model. We propose a metric that computes this proximity based on three phenomena of NC: within-class variability collapse, simplex encoded label interpolation geometry structure is formed, and the nearest center classifier becomes optimal on training data. Extensive experiments on 11 benchmark datasets demonstrate the effectiveness and efficiency of the proposed method over the existing SOTA approaches. Particularly, our method achieves SOTA transferability estimation accuracy with approximately 10xwall-clock time speed up compared to the existing approaches

Xudong Wang, Li Lyna Zhang, Jiahang Xu, Quanlu Zhang, Yujing Wang, Yuqing Yang, Ningxin Zheng, Ting Cao, Mao Yang

The combination of Neural Architecture Search (NAS) and quantization has proven successful in automatically designing low-FLOPs INT8 quantized neural networks (QNN). However, directly applying NAS to design accurate QNN models that achieve low latency on real-world devices leads to inferior performance. In this work, we identify that the poor INT8 latency is due to the quantization-unfriendly issue: the operator and configuration (e.g., channel width) choices in prior art search spaces lead to diverse quantization efficiency and can slow down the INT8 inference speed. To address this challenge, we propose SpaceEvo, an automatic method for designing a dedicated, quantization-friendly search space for each target hardware. The key idea of SpaceEvo is to automatically search hardware-preferred operators and configurations to construct the search space, guided by a metric called Q-T score to quantify how quantization-friendly a candidate search space is. We further train a quantized-for-all supernet over our discovered search space, enabling the searched models to be directly deployed without extra retraining or quantization. Our discovered models, SEQnet, establish new SOTA INT8 quantized accuracy under various latency constraints, achieving up to 10.1% accuracy improvement on ImageNet than prior art CNNs under the same latency. Extensive experiments on real devices show that SpaceEvo consistently outperforms manually-designed search spaces with up to 2.5x faster speed while achieving the same accuracy.

Evonne Ng, Sanjay Subramanian, Dan Klein, Angjoo Kanazawa, Trevor Darrell, Shiry Ginosar

We present a framework for generating appropriate facial responses from a listener in dyadic social interactions based on the speaker's words. Given an input transcription of the speaker's words with their timestamps, our approach autoregressively predicts a response of a listener: a sequence of listener facial gestures, quantized using a VQ-VAE. Since gesture is a language component, we propose treating the quantized atomic motion elements as additional language token inputs to a transformer-based large language model. Initializing our transformer with the weights of a language model pre-trained only on text results in significantly higher quality listener responses than training a transformer from scratch. We show that our generated listener motion is fluent and reflective of language semantics through quantitative metrics and a qualitative user study. In our evaluation, we analyze the model's ability to utilize temporal and semantic aspects of spoken text.

Anqi Yang, Eunhee Kang, Hyong-Euk Lee, Aswin C. Sankaranarayanan

Diffractive blur and low light levels are two fundamental challenges in producing high-quality photographs in under-display cameras (UDCs). In this paper, we incorporate phase masks on display panels to tackle both challenges. Our design inserts two phase masks, specifically two microlens arrays, in front of and behind a display panel. The first phase mask concentrates light on the locations where the display is transparent so that more light passes through the display, and the second phase mask reverts the effect of the first phase mask. We further optimize the folding height of each microlens to improve the quality of PSFs and suppress chromatic aberration. We evaluate our design using a physically-accurate simulator based on Fourier optics. The proposed design is able to double the light throughput while improving the invertibility of the PSFs. Lastly, we discuss the effect of our design on the display quality and show that implementation with polarization-dependent phase masks can leave the display quality uncompromised.

Chaoran Tian, Weihong Pan, Zimo Wang, Mao Mao, Guofeng Zhang, Hujun Bao, Ping Tan, Zhaopeng Cui

Stereo depth estimation usually struggles to deal with textureless scenes for both traditional and learning-based methods due to the inherent dependence on image correspondence matching. In this paper, we propose a novel neural network, i.e., DPS-Net, to exploit both the prior geometric knowledge and polarimetric information for depth estimation with two polarimetric stereo images. Specifically, we construct both RGB and polarization correlation volumes to fully leverage the multi-domain similarity between polarimetric stereo images. Since inherent ambiguities exist in the polarization images, we introduce the iso-depth cost explicitly into the network to solve these ambiguities. Moreover, we design a cascaded dual-GRU architecture to recurrently update the disparity and effectively fuse both the multi-domain correlation features and the iso-depth cost. Besides, we present new synthetic and real polarimetric stereo datasets for evaluation. Experimental results demonstrate that our method outperforms the state-of-the-art stereo depth estimation methods.

Kai Liu, Sheng Jin, Zhihang Fu, Ze Chen, Rongxin Jiang, Jieping Ye

Without manually annotated identities, unsupervised multi-object trackers are inferior to learning reliable feature embeddings. It causes the similarity-based inter-frame association stage also be error-prone, where an uncertainty problem arises. The frame-by-frame accumulated uncertainty prevents trackers from learning the consistent feature embedding against time variation. To avoid this uncertainty problem, recent self-supervised techniques are adopted, whereas they failed to capture temporal relations. The inter-frame uncertainty still exists. In fact, this paper argues that though the uncertainty problem is inevitable, it is possible to leverage the uncertainty itself to improve the learned consistency in turn. Specifically, an uncertainty-based metric is developed to verify and rectify the risky associations. The resulting accurate pseudo-tracklets boost learning the feature consistency. And accurate tracklets can incorporate temporal information into spatial transformation. This paper proposes a tracklet-guided augmentation strategy to simulate the tracklet's motion, which adopts a hierarchical uncertainty-based sampling mechanism for hard sample mining. The ultimate unsupervised MOT framework, namely U2MOT, is proven effective on MOT-Challenges and VisDrone-MOT benchmark. U2MOT achieves a SOTA performance among the published supervised and unsupervised trackers.

Seong Min Kye, Kwanghee Choi, Hyeongmin Byun, Buru Chang

Active learning (AL) aims to select the most useful data samples from an unlabeled data pool and annotate them to expand the labeled dataset under a limited budget.Especially, uncertainty-based methods choose the most uncertain samples, which are known to be effective in improving model performance.However, AL literature often overlooks training dynamics (TD), defined as the ever-changing model behavior during optimization via stochastic gradient descent, even though other research areas have empirically shown that TD provides important clues for measuring the data uncertainty. In this paper, we first provide theoretical and empirical evidence to argue the usefulness of utilizing the ever-changing model behavior rather than the fully trained model snapshot. We then propose a novel AL method, Training Dynamics for Active Learning (TiDAL), which efficiently predicts the training dynamics of unlabeled data to estimate their uncertainty. Experimental results show that our TiDAL achieves better or comparable performance on both balanced and imbalanced benchmark datasets compared to state-of-the-art AL methods, which estimate data uncertainty using only static information after model training.

Shuai Yuan, Shuzhi Yu, Hannah Kim, Carlo Tomasi

Unsupervised optical flow estimation is especially hard near occlusions and motion boundaries and in low-texture regions. We show that additional information such as semantics and domain knowledge can help better constrain this problem. We introduce SemARFlow, an unsupervised optical flow network designed for autonomous driving data that takes estimated semantic segmentation masks as additional inputs. This additional information is injected into the encoder and into a learned upsampler that refines the flow output. In addition, a simple yet effective semantic augmentation module provides self-supervision when learning flow and its boundaries for vehicles, poles, and sky. Together, these injections of semantic information improve the KITTI-2015 optical flow test error rate from 11.80% to 8.38%. We also show visible improvements around object boundaries as well as a greater ability to generalize across datasets. Code is available at https://github.com/duke-vision/semantic-unsup-flow-release.

Nikolaos-Antonios Ypsilantis, Kaifeng Chen, Bingyi Cao, Mário Lipovský, Pelin Dogan-Schönberger, Grzegorz Makosa, Boris Bluntschli, Mojtaba Seyedhosseini, Ondřej Chum, André Araujo

Fine-grained and instance-level recognition methods are commonly trained and evaluated on specific domains, in a model per domain scenario. Such an approach, however, is impractical in real large-scale applications. In this work, we address the problem of universal image embedding, where a single universal model is trained and used in multiple domains. First, we leverage existing domain-specific datasets to carefully construct a new large-scale public benchmark for the evaluation of universal image embeddings, with 241k query images, 1.4M index images and 2.8M training images across 8 different domains and 349k classes. We define suitable metrics, training and evaluation protocols to foster future research in this area. Second, we provide a comprehensive experimental evaluation on the new dataset, demonstrating that existing approaches and simplistic extensions lead to worse performance than an assembly of models trained for each domain separately. Finally, we conducted a public research competition on this topic, leveraging industrial datasets, which attracted the participation of more than 1k teams worldwide. This exercise generated many interesting research ideas and findings which we present in detail. Project webpage: https://cmp.felk.cvut.cz/univ_emb/

Yiming Xie, Huaizu Jiang, Georgia Gkioxari, Julian Straub

We present PARQ - a multi-view 3D object detector with transformer and pixel-aligned recurrent queries. Unlike previous works that use learnable features or only encode 3D point positions as queries in the decoder, PARQ leverages appearance-enhanced queries initialized from reference points in 3D space and updates their 3D location with recurrent cross-attention operations. Incorporating pixel-aligned features and cross attention enables the model to encode the necessary 3D-to-2D correspondences and capture global contextual information of the input images. PARQ outperforms prior best methods on the ScanNet and ARKitScenes datasets, learns and detects faster, is more robust to distribution shifts in reference points, can leverage additional input views without retraining, and can adapt inference compute by changing the number of recurrent iterations.