ICCV 2021 论文列表

Neural implicit 3D representations have emerged as a powerful paradigm for reconstructing surfaces from multi-view images and synthesizing novel views. Unfortunately, existing methods such as DVR or IDR require accurate per-pixel object masks as supervision. At the same time, neural radiance fields have revolutionized novel view synthesis. However, NeRF's estimated volume density does not admit accurate surface reconstruction. Our key insight is that implicit surface models and radiance fields can be formulated in a unified way, enabling both surface and volume rendering using the same model. This unified perspective enables novel, more efficient sampling procedures and the ability to reconstruct accurate surfaces without input masks. We compare our method on the DTU, BlendedMVS, and a synthetic indoor dataset. Our experiments demonstrate that we outperform NeRF in terms of reconstruction quality while performing on par with IDR without requiring masks.

Recent studies have shown that cascade cost volume can play a vital role in deep stereo matching to achieve high resolution depth map with efficient hardware usage. However, how to construct good cascade volume as well as effective sampling for them are still under in-depth study. Previous cascade-based methods usually perform uniform sampling in a predicted disparity range based on variance, which easily misses the ground truth disparity and decreases disparity map accuracy. In this paper, we propose an uncertainty adaptive sampling network (UASNet) featuring two modules: an uncertainty distribution-guided range prediction (URP) model and an uncertainty-based disparity sampler (UDS) module. The URP explores the more discriminative uncertainty distribution to handle the complex matching ambiguities and to improve disparity range prediction. The UDS adaptively adjusts sampling interval to localize disparity with improved accuracy. With the proposed modules, our UASNet learns to construct cascade cost volume and predict full-resolution disparity map directly. Extensive experiments show that the proposed method achieves the highest ground truth covering ratio compared with other cascade cost volume based stereo matching methods. Our method also achieves top performance on both SceneFlow dataset and KITTI benchmark.

Unsupervised domain adaptation (UDA) methods can dramatically improve generalization on unlabeled target domains. However, optimal hyper-parameter selection is critical to achieving high accuracy and avoiding negative transfer. Supervised hyper-parameter validation is not possible without labeled target data, which raises the question: How can we validate unsupervised adaptation techniques in a realistic way? We first empirically analyze existing criteria and demonstrate that they are not very effective for tuning hyper-parameters. Intuitively, a well-trained source classifier should embed target samples of the same class nearby, forming dense neighborhoods in feature space. Based on this assumption, we propose a novel unsupervised validation criterion that measures the density of soft neighborhoods by computing the entropy of the similarity distribution between points. Our criterion is simpler than competing validation methods, yet more effective; it can tune hyper-parameters and the number of training iterations in both image classification and semantic segmentation models.

Automatic augmentation methods have recently become a crucial pillar for strong model performance in vision tasks. While existing automatic augmentation methods need to trade off simplicity, cost and performance, we present a most simple baseline, TrivialAugment, that outperforms previous methods for almost free. TrivialAugment is parameter-free and only applies a single augmentation to each image. Thus, TrivialAugment's effectiveness is very unexpected to us and we performed very thorough experiments to study its performance. First, we compare TrivialAugment to previous state-of-the-art methods in a variety of image classification scenarios. Then, we perform multiple ablation studies with different augmentation spaces, augmentation methods and setups to understand the crucial requirements for its performance. Additionally, we provide a simple interface to facilitate the widespread adoption of automatic augmentation methods, as well as our full code base for reproducibility. Since our work reveals a stagnation in many parts of automatic augmentation research, we end with a short proposal of best practices for sustained future progress in automatic augmentation methods.

With the development of deep convolutional neural networks, image matting has ushered in a new phase. Regarding the nature of image matting, most researches have focused on solutions for transition regions. However, we argue that many existing approaches are excessively focused on transition-dominant local fields and ignored the inherent coordination between global information and transition optimisation. In this paper, we propose the Tripartite Information Mining and Integration Network (TIMI-Net) to harmonize the coordination between global and local attributes formally. Specifically, we resort to a novel 3-branch encoder to accomplish comprehensive mining of the input information, which can supplement the neglected coordination between global and local fields. In order to achieve effective and complete interaction between such multi-branches information, we develop the Tripartite Information Integration (TI^2) Module to transform and integrate the interconnections between the different branches. In addition, we built a large-scale human matting dataset (Human-2K) to advance human image matting, which consists of 2100 high-precision human images (2000 images for training and 100 images for test). Finally, we conduct extensive experiments to prove the performance of our proposed TIMI-Net, which demonstrates that our method performs favourably against the SOTA approaches on the alphamatting.com (Rank First), Composition-1K (MSE-0.006, Grad-11.5), Distinctions-646 and our Human-2K. Also, we have developed an online evaluation website to perform natural image matting. Project page: https://wukaoliu.github.io/TIMI-Net.

In this paper, we tackle the detection of out-of-distribution (OOD) objects in semantic segmentation. By analyzing the literature, we found that current methods are either accurate or fast but not both which limits their usability in real world applications. To get the best of both aspects, we propose to mitigate the common shortcomings by following four design principles: decoupling the OOD detection from the segmentation task, observing the entire segmentation network instead of just its output, generating training data for the OOD detector by leveraging blind spots in the segmentation network and focusing the generated data on localized regions in the image to simulate OOD objects. Our main contribution is a new OOD detection architecture called ObsNet associated with a dedicated training scheme based on Local Adversarial Attacks (LAA). We validate the soundness of our approach across numerous ablation studies. We also show it obtains top performances both in speed and accuracy when compared to ten recent methods of the literature on three different datasets.

This paper aims to reconstruct physically plausible hand motion from monocular color images. Existing frame-by-frame estimating approaches can not guarantee the physical plausibility (e.g. penetration, jittering) directly. In this paper, we embed physical constraints on the per-frame estimated motions in both spatial and temporal space. Our key idea is to adopt a self-supervised learning strategy to train a novel encoder-decoder, named TravelNet, whose training motion data is prepared by the physics engine using discrete pose states. TravelNet captures key pose states from hand motion sequences as compact motion descriptors, inspired by the concept of keyframes in animation. Finally, it manages to extract those key states out of perturbations without manual annotations, and reconstruct the motions preserving details and physical plausibility. In the experiments, we show that the outputs of the TravelNet contain both finger synergism and time consistency. Through the proposed framework, hand motions can be accurately reconstructed and flexibly re-edited, which is superior to the state-of-the-art methods.

Open-set semi-supervised learning (open-set SSL) investigates a challenging but practical scenario where out-of-distribution (OOD) samples are contained in the unlabeled data. While the mainstream technique seeks to completely filter out the OOD samples for semi-supervised learning (SSL), we propose a novel training mechanism that could effectively exploit the presence of OOD data for enhanced feature learning while avoiding its adverse impact on the SSL. We achieve this goal by first introducing a warm-up training that leverages all the unlabeled data, including both the in-distribution (ID) and OOD samples. Specifically, we perform a pretext task that enforces our feature extractor to obtain a high-level semantic understanding of the training images, leading to more discriminative features that can benefit the downstream tasks. Since the OOD samples are inevitably detrimental to SSL, we propose a novel cross-modal matching strategy to detect OOD samples. Instead of directly applying binary classification, we train the network to predict whether the data sample is matched to an assigned one-hot class label. The appeal of the proposed cross-modal matching over binary classification is the ability to generate a compatible feature space that aligns with the core classification task. Extensive experiments show that our approach substantially lifts the performance on open-set SSL and outperforms the state-of-the-art by a large margin.

Existing Unsupervised Domain Adaptation (UDA) literature adopts the covariate shift and conditional shift assumptions, which essentially encourage models to learn common features across domains. However, due to the lack of supervision in the target domain, they suffer from the semantic loss: the feature will inevitably lose non-discriminative semantics in source domain, which is however discriminative in target domain. We use a causal view---transportability theory---to identify that such loss is in fact a confounding effect, which can only be removed by causal intervention. However, the theoretical solution provided by transportability is far from practical for UDA, because it requires the stratification and representation of the unobserved confounder that is the cause of the domain gap. To this end, we propose a practical solution: Transporting Causal Mechanisms (TCM), to identify the confounder stratum and representations by using the domain-invariant disentangled causal mechanisms, which are discovered in an unsupervised fashion. Our TCM is both theoretically and empirically grounded. Extensive experiments show that TCM achieves state-of-the-art performance on three challenging UDA benchmarks: ImageCLEF-DA, Office-Home, and VisDA-2017. Codes are available at https://github.com/yue-zhongqi/tcm.

Visual tracking has achieved considerable progress in recent years. However, current research in the field mainly focuses on tracking of opaque objects, while little attention is paid to transparent object tracking. In this paper, we make the first attempt in exploring this problem by proposing a Transparent Object Tracking Benchmark (TOTB). Specifically, TOTB consists of 225 videos (86K frames) from 15 diverse transparent object categories. Each sequence is manually labeled with axis-aligned bounding boxes. To the best of our knowledge, TOTB is the first benchmark dedicated to transparent object tracking. In order to understand how existing trackers perform and to provide comparison for future research on TOTB, we extensively evaluate 25 state-of-the-art tracking algorithms. The evaluation results exhibit that more efforts are needed to improve transparent object tracking. Besides, we observe some nontrivial findings from the evaluation that are discrepant with some common beliefs in opaque object tracking. For example, we find that deep(er) features are not always good for improvements. Moreover, to encourage future research, we introduce a novel tracker, named TransATOM, which leverages transparency features for tracking and surpasses all 25 evaluated approaches by a large margin. By releasing TOTB, we expect to facilitate future research and application of transparent object tracking in both the academia and industry. The TOTB and evaluation results as well as TransATOM are available at https://hengfan2010.github.io/projects/TOTB/.

Recently RGB-D sensors have become very popular in the area of Simultaneous Localisation and Mapping (SLAM). The RGB-D SLAM approach relies heavily on the accuracy of the input depth map. However, refraction and reflection of transparent objects will result in false depth input of RGB-D cameras, which makes the traditional RGB-D SLAM algorithm unable to work correctly in the presence of transparent objects. In this paper, we propose a novel SLAM approach called transfusion that allows transparent object existence and recovery in the video input. Our method is composed of two parts. Transparent Objects Cut Iterative Closest Points (TC-ICP)is first used to recover camera pose, detecting and removing transparent objects from input to reduce the trajectory errors. Then Transparent Objects Reconstruction (TO-Reconstruction) is used to reconstruct the transparent objects and opaque objects separately. The opaque objects are reconstructed with the traditional method, and the transparent objects are reconstructed with the visual hull-based method. To evaluate our algorithm, we construct a new RGB-D SLAM database containing 25 video sequences. Each sequence has at least one transparent object. Experiments show that our approach can work adequately in scenes contain transparent objects while the existing approach can not handle them. Our approach significantly improves the accuracy of the camera trajectory and the quality of environment reconstruction.

This work studies the Tensor Robust Principal Component Analysis (TRPCA) problem, which aims to exactly recover the low-rank and sparse components from their sum. Our model is motivated by the recently proposed linear transforms based tensor-tensor product and tensor SVD. We define a new transforms depended tensor rank and the corresponding tensor nuclear norm. Then we solve the TRPCA problem by convex optimization whose objective is a weighted combination of the new tensor nuclear norm and l_1-norm. In theory, we prove that under some incoherence conditions, the convex program exactly recovers the underlying low-rank and sparse components with high probability. Our new TRPCA is much more general since it allows to use any invertible linear transforms. Thus, we have more choices in practice for different tasks and different type of data. Numerical experiments verify our results and the application on image recovery demonstrates the superiority of our method.

The simultaneous recognition of multiple objects in one image remains a challenging task, spanning multiple events in the recognition field such as various object scales, inconsistent appearances, and confused inter-class relationships. Recent research efforts mainly resort to the statistic label co-occurrences and linguistic word embedding to enhance the unclear semantics. Different from these researches, in this paper, we propose a novel Transformer-based Dual Relation learning framework, constructing complementary relationships by exploring two aspects of correlation, i.e., structural relation graph and semantic relation graph. The structural relation graph aims to capture long-range correlations from object context, by developing a cross-scale transformer-based architecture. The semantic graph dynamically models the semantic meanings of image objects with explicit semantic-aware constraints. In addition, we also incorporate the learnt structural relationship into the semantic graph, constructing a joint relation graph for robust representations. With the collaborative learning of these two effective relation graphs, our approach achieves new state-of-the-art on two popular multi-label recognition benchmarks, i.e. MS-COCO and VOC 2007 dataset.

While convolutional neural networks have shown a tremendous impact on various computer vision tasks, they generally demonstrate limitations in explicitly modeling long-range dependencies due to the intrinsic locality of the convolution operation. Initially designed for natural language processing tasks, Transformers have emerged as alternative architectures with innate global self-attention mechanisms to capture long-range dependencies. In this paper, we propose TransDepth, an architecture that benefits from both convolutional neural networks and transformers. To avoid the network losing its ability to capture local-level details due to the adoption of transformers, we propose a novel decoder that employs attention mechanisms based on gates. Notably, this is the first paper that applies transformers to pixel-wise prediction problems involving continuous labels (i.e., monocular depth prediction and surface normal estimation). Extensive experiments demonstrate that the proposed TransDepth achieves state-of-the-art performance on three challenging datasets. Our code is available at: https://github.com/ygjwd12345/TransDepth.

Image-to-image (I2I) translation has matured in recent years and is able to generate high-quality realistic images. However, despite current success, it still faces important challenges when applied to small domains. Existing methods use transfer learning for I2I translation, but they still require the learning of millions of parameters from scratch. This drawback severely limits its application on small domains. In this paper, we propose a new transfer learning for I2I translation (TransferI2I). We decouple our learning process into the image generation step and the I2I translation step. In the first step we propose two novel techniques: source-target initialization and self-initialization of the adaptor layer. The former finetunes the pretrained generative model (e.g., StyleGAN) on source and target data. The latter allows to initialize all non-pretrained network parameters without the need of any data. These techniques provide a better initialization for the I2I translation. Second step performs the actual I2I translation using the learned weights in the first step. In addition, we introduce an auxiliary GAN that further facilitates the training of deep I2I systems even from small datasets. In extensive experiments on three datasets, (Animal faces, Birds, and Foods), we show that we outperform existing methods and that mFID improves on several datasets with over 25 points.

Few-shot learning (FSL) attempts to learn with limited data. In this work, we perform the feature extraction in the Euclidean space and the geodesic distance metric on the Oblique Manifold (OM). Specially, for better feature extraction, we propose a non-parametric Region Self-attention with Spatial Pyramid Pooling (RSSPP), which realizes a trade-off between the generalization and the discriminative ability of the single image feature. Then, we embed the feature to OM as a point. Furthermore, we design an Oblique Distance-based Classifier (ODC) that achieves classification in the tangent spaces which better approximate OM locally by learnable tangency points. Finally, we introduce a new method for parameters initialization and a novel loss function in the transductive settings. Extensive experiments demonstrate the effectiveness of our algorithm and it outperforms state-of-the-art methods on the popular benchmarks: mini-ImageNet, tiered-ImageNet, and Caltech-UCSD Birds-200-2011 (CUB).

We show that relation modeling between visual elements matters in cropping view recommendation. Cropping view recommendation addresses the problem of image recomposition conditioned on the composition quality and the ranking of views (cropped sub-regions). This task is challenging because the visual difference is subtle when a visual element is reserved or removed. Existing methods represent visual elements by extracting region-based convolutional features inside and outside the cropping view boundaries, without probing a fundamental question: why some visual elements are of interest or of discard? In this work, we observe that the relation between different visual elements significantly affects their relative positions to the desired cropping view, and such relation can be characterized by the attraction inside/outside the cropping view boundaries and the repulsion across the boundaries. By instantiating a transformer-based solution that represents visual elements as visual words and that models the dependencies between visual words, we report not only state of-the-art performance on public benchmarks, but also interesting visualizations that depict the attraction and repulsion between visual elements, which may shed light on what makes for effective cropping view recommendation.

In this paper, we present a neat yet effective transformer-based framework for visual grounding, namely TransVG, to address the task of grounding a language query to the corresponding region onto an image. The state-of-the-art methods, including two-stage or one-stage ones, rely on a complex module with manually-designed mechanisms to perform the query reasoning and multi-modal fusion. However, the involvement of certain mechanisms in fusion module design, such as query decomposition and image scene graph, makes the models easily overfit to datasets with specific scenarios, and limits the plenitudinous interaction between the visual-linguistic context. To avoid this caveat, we propose to establish the multi-modal correspondence by leveraging transformers, and empirically show that the complex fusion modules (e.g., modular attention network, dynamic graph, and multi-modal tree) can be replaced by a simple stack of transformer encoder layers with higher performance. Moreover, we re-formulate the visual grounding as a direct coordinates regression problem and avoid making predictions out of a set of candidates (i.e., region proposals or anchor boxes). Extensive experiments are conducted on five widely used datasets, and a series of state-of-the-art records are set by our TransVG. We build the benchmark of transformer-based visual grounding framework and make the code available at https://github.com/djiajunustc/TransVG.

Extracting robust feature representation is one of the key challenges in object re-identification (ReID). Although convolution neural network (CNN)-based methods have achieved great success, they only process one local neighborhood at a time and suffer from information loss on details caused by convolution and downsampling operators pooling and strided convolution).To overcome these limitations, we propose a pure transformer-based object ReID framework named TransReID. Specifically, we first encode an image as a sequence of patches and build a transformer-based strong baseline with a few critical improvements, which achieves competitive results on several ReID benchmarks with CNN-based methods. To further enhance the robust feature learning in the context of transformers, two novel modules are carefully designed. (i) The jigsaw patch module (JPM) is proposed to rearrange the patch embeddings via shift and patch shuffle operations which generates robust features with improved discrimination ability and more diversified coverage. (ii) The side information embeddings (SIE) is introduced to mitigate feature bias towards camera/view variations by plugging in learnable embeddings to incorporate these non-visual clues. To the best of our knowledge, this is the first work to adopt a pure transformer for ReID research. Experimental results of TransReID are superior promising, which achieve state-of-the-art performance on both person and vehicle ReID benchmarks. Code is available at https://github.com/heshuting555/TransReID