Advanced self-supervised visual representation learning methods rely on the instance discrimination (ID) pretext task. We point out that the ID task has an implicit semantic consistency (SC) assumption, which may not hold in unconstrained datasets. In this paper, we propose a novel contrastive mask prediction (CMP) task for visual representation learning and design a mask contrast (MaskCo) framework to implement the idea. MaskCo contrasts region-level features instead of view-level features, which makes it possible to identify the positive sample without any assumptions. To solve the domain gap between masked and unmasked features, we design a dedicated mask prediction head in MaskCo. This module is shown to be the key to the success of the CMP. We evaluated MaskCo on training datasets beyond ImageNet and compare its performance with MoCo V2. Results show that MaskCo achieves comparable performance with MoCo V2 using ImageNet training dataset, but demonstrates a stronger performance across a range of downstream tasks when COCO or Conceptual Captions are used for training. MaskCo provides a promising alternative to the ID-based methods for self-supervised learning in the wild.
论文检索
输入标题、作者或关键词，从 6,469 篇学术成果中精准定位
In semantic segmentation tasks, input images can often have more than one plausible interpretation, thus allowing for multiple valid labels. To capture such ambiguities, recent work has explored the use of probabilistic networks that can learn a distribution over predictions. However, these do not necessarily represent the empirical distribution accurately. In this work, we present a strategy for learning a calibrated predictive distribution over semantic maps, where the probability associated with each prediction reflects its ground truth correctness likelihood. To this end, we propose a novel two-stage, cascaded approach for calibrated adversarial refinement: (i) a standard segmentation network is trained with categorical cross-entropy to predict a pixelwise probability distribution over semantic classes and (ii) an adversarially trained stochastic network is used to model the inter-pixel correlations to refine the output of the first network into coherent samples. Importantly, to calibrate the refinement network and prevent mode collapse, the expectation of the samples in the second stage is matched to the probabilities predicted in the first. We demonstrate the versatility and robustness of the approach by achieving state-of-the-art results on the multigrader LIDC dataset and on a modified Cityscapes dataset with injected ambiguities. In addition, we show that the core design can be adapted to other tasks requiring learning a calibrated predictive distribution by experimenting on a toy regression dataset. We provide an open source implementation of our method at https://github.com/EliasKassapis/CARSSS.
Metric learning has received conflicting assessments concerning its suitability for solving instance segmentation tasks. It has been dismissed as theoretically flawed due to the shift equivariance of the employed CNNs and their respective inability to distinguish same-looking objects. Yet it has been shown to yield state of the art results for a variety of tasks, and practical issues have mainly been reported in the context of tile-and-stitch approaches, where discontinuities at tile boundaries have been observed. To date, neither of the reported issues have undergone thorough formal analysis. In our work, we contribute a comprehensive formal analysis of the shift equivariance properties of encoder-decoder-style CNNs, which yields a clear picture of what can and cannot be achieved with metric learning in the face of same-looking objects. In particular, we prove that a standard encoder-decoder network that takes d-dimensional images as input, with l pooling layers and pooling factor f, has the capacity to distinguish at most f^(dl) same-looking objects, and we show that this upper limit can be reached. Furthermore, we show that to avoid discontinuities in a tile-and-stitch approach, assuming standard batch size 1, it is necessary to employ valid convolutions in combination with a training output window size strictly greater than f^l, while at test-time it is necessary to crop tiles to size n * f^l before stitching, with n >= 1. We complement these theoretical findings by discussing a number of insightful special cases for which we show empirical results on synthetic and real data.
This paper presents a novel task together with a new benchmark for detecting generic, taxonomy-free event boundaries that segment a whole video into chunks. Conventional work in temporal video segmentation and action detection focuses on localizing pre-defined action categories and thus does not scale to generic videos. Cognitive Science has known since last century that humans consistently segment videos into meaningful temporal chunks. This segmentation happens naturally, without pre-defined event categories and without being explicitly asked to do so. Here, we repeat these cognitive experiments on mainstream CV datasets; with our novel annotation guideline which addresses the complexities of taxonomy-free event boundary annotation, we introduce the task of Generic Event Boundary Detection (GEBD) and the new benchmark Kinetics-GEBD. We view GEBD as an important stepping stone towards understanding the video as a whole, and believe it has been previously neglected due to a lack of proper task definition and annotations. Through experiment and human study we demonstrate the value of the annotations. Further, we benchmark supervised and un-supervised GEBD approaches on the TAPOS dataset and our Kinetics-GEBD. We release our annotations and baseline codes at CVPR'21 LOVEU Challenge: https://sites.google.com/view/loveucvpr21.
Reconstructing 3D shape from 2D sketches has long been an open problem because the sketches only provide very sparse and ambiguous information. In this paper, we use an encoder/decoder architecture for the sketch to mesh translation. When integrated into a user interface that provides camera parameters for the sketches, this enables us to leverage its latent parametrization to represent and refine a 3D mesh so that its projections match the external contours outlined in the sketch. We will show that this approach is easy to deploy, robust to style changes, and effective. Furthermore, it can be used for shape refinement given only single pen strokes. We compare our approach to state-of-the-art methods on sketches - both hand-drawn and synthesized - and demonstrate that we outperform them.
Video content creation keeps growing at an incredible pace; yet, creating engaging stories remains challenging and requires non-trivial video editing expertise. Many video editing components are astonishingly hard to automate primarily due to the lack of raw video materials. This paper focuses on a new task for computational video editing, namely the task of raking cut plausibility. Our key idea is to leverage content that has already been edited to learn fine-grained audiovisual patterns that trigger cuts. To do this, we first collected a data source of more than 10K videos, from which we extract more than 260K cuts. We devise a model that learns to discriminate between real and artificial cuts via contrastive learning. We set up a new task and a set of baselines to benchmark video cut generation. We observe that our proposed model outperforms the baselines by large margins. To demonstrate our model in real-world applications, we conduct human studies in a collection of unedited videos. The results show that our model does a better job at cutting than random and alternative baselines.
Non-Rigid Neural Radiance Fields: Reconstruction and Novel View Synthesis of a Dynamic Scene From Monocular Video
PDF ↗We present Non-Rigid Neural Radiance Fields (NR-NeRF), a reconstruction and novel view synthesis approach for general non-rigid dynamic scenes. Our approach takes RGB images of a dynamic scene as input (e.g., from a monocular video recording), and creates a high-quality space-time geometry and appearance representation. We show that a single handheld consumer-grade camera is sufficient to synthesize sophisticated renderings of a dynamic scene from novel virtual camera views, e.g. a `bullet-time' video effect. NR-NeRF disentangles the dynamic scene into a canonical volume and its deformation. Scene deformation is implemented as ray bending, where straight rays are deformed non-rigidly. We also propose a novel rigidity network to better constrain rigid regions of the scene, leading to more stable results. The ray bending and rigidity network are trained without explicit supervision. Our formulation enables dense correspondence estimation across views and time, and compelling video editing applications such as motion exaggeration. Our code will be open sourced.
Recent advancements in differentiable rendering and 3D reasoning have driven exciting results in novel view synthesis from a single image. Despite realistic results, methods are limited to relatively small view change. In order to synthesize immersive scenes, models must also be able to extrapolate. We present an approach that fuses 3D reasoning with autoregressive modeling to outpaint large view changes in a 3D-consistent manner, which enables scene synthesis. We demonstrate considerable improvement in single-image large-angle view synthesis results compared to a variety of methods and possible variants across simulated and real datasets. In addition, we show increased 3D consistency compared to alternative accumulation methods.
There is an emerging sense that the vulnerability of Image Convolutional Neural Networks (CNN), i.e., sensitivity to image corruptions, perturbations, and adversarial attacks, is connected with Texture Bias. This relative lack of Shape Bias is also responsible for poor performance in Domain Generalization (DG). The inclusion of a role of shape alleviates these vulnerabilities and some approaches have achieved this by training on negative images, images endowed with edge maps, or images with conflicting shape and texture information. This paper advocates an explicit and complete representation of shape using a classical computer vision approach, namely, representing the shape content of an image with the shock graph of its contour map. The resulting graph and its descriptor is a complete representation of contour content and is classified using recent Graph Neural Network (GNN) methods. The experimental results on three domain shift datasets, Colored MNIST, PACS, and VLCS demonstrate that even without using appearance the shape-based approach exceeds classical Image CNN based methods in domain generalization.
We address the problem of network quantization, that is, reducing bit-widths of weights and/or activations to lighten network architectures. Quantization methods use a rounding function to map full-precision values to the nearest quantized ones, but this operation is not differentiable. There are mainly two approaches to training quantized networks with gradient-based optimizers. First, a straight-through estimator (STE) replaces the zero derivative of the rounding with that of an identity function, which causes a gradient mismatch problem. Second, soft quantizers approximate the rounding with continuous functions at training time, and exploit the rounding for quantization at test time. This alleviates the gradient mismatch, but causes a quantizer gap problem. We alleviate both problems in a unified framework. To this end, we introduce a novel quantizer, dubbed a distance-aware quantizer (DAQ), that mainly consists of a distance-aware soft rounding (DASR) and a temperature controller. To alleviate the gradient mismatch problem, DASR approximates the discrete rounding with the kernel soft argmax, which is based on our insight that the quantization can be formulated as a distance-based assignment problem between full-precision values and quantized ones. The controller adjusts the temperature parameter in DASR adaptively according to the input, addressing the quantizer gap problem. Experimental results on standard benchmarks show that DAQ outperforms the state of the art significantly for various bit-widths without bells and whistles.
The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. It is hypothesized that the well known tendency of standard classifier training to overfit to popular classes can be exploited for effective transfer learning. Rather than eliminating this overfitting, e.g. by adopting popular class-balanced sampling methods, the learning algorithm should instead leverage this overfitting to transfer geometric information from popular to low-shot classes. A new classifier architecture, GistNet, is proposed to support this goal, using constellations of classifier parameters to encode the class geometry. A new learning algorithm is then proposed for GeometrIc Structure Transfer (GIST), with resort to a combination of loss functions that combine class-balanced and random sampling to guarantee that, while overfitting to the popular classes is restricted to geometric parameters, it is leveraged to transfer class geometry from popular to few-shot classes. This enables better generalization for few-shot classes without the need for the manual specification of class weights, or even the explicit grouping of classes into different types. Experiments on two popular long-tailed recognition datasets show that GistNet outperforms existing solutions to this problem.
Deep neural networks often make decisions based on the spurious correlations inherent in the dataset, failing to generalize in an unbiased data distribution. Although previous approaches pre-define the type of dataset bias to prevent the network from learning it, recognizing the bias type in the real dataset is often prohibitive. This paper proposes a novel bias-tailored augmentation-based approach, BiaSwap, for learning debiased representation without requiring supervision on the bias type. Motivated by the phenomenon that the bias corresponds to the attributes the model learns as a shortcut, we utilize an image-to-image translation model optimized to transfer the attributes that the classifier often learns easily. As a prerequisite, we sort the training samples based on how much a biased model exploits them as a shortcut and divide them into bias-guiding and bias-contrary samples in an unsupervised manner. Afterwards, we utilize the CAM of GCE-trained classifier in the patch cooccurrence discriminator in order to focus on translating the bias attributes. Therefore, given the pair of bias-guiding and bias-contrary, the model generates the augmented bias-swapped image which contains the bias attributes from the bias-contrary images, while preserving bias-irrelevant ones in the bias-guiding images. We demonstrate the superiority of our approach against the baselines over both synthetic and real-world datasets. Even without careful supervision on the bias, BiaSwap achieves a remarkable performance on both unbiased and bias-guiding samples, implying the improved generalization capability of the model.
Finding local features that are repeatable across multiple views is a cornerstone of sparse 3D reconstruction. The classical image matching paradigm detects keypoints per-image once and for all, which can yield poorly-localized features and propagate large errors to the final geometry. In this paper, we refine two key steps of structure-from-motion by a direct alignment of low-level image information from multiple views: we first adjust the initial keypoint locations prior to any geometric estimation, and subsequently refine points and camera poses as a post-processing. This refinement is robust to large detection noise and appearance changes, as it optimizes a featuremetric error based on dense features predicted by a neural network. This significantly improves the accuracy of camera poses and scene geometry for a wide range of keypoint detectors, challenging viewing conditions, and off-the-shelf deep features. Our system easily scales to large image collections, enabling pixel-perfect crowd-sourced localization at scale. Our code is publicly available at https://github.com/cvg/pixel-perfect-sfm as an add-on to the popular SfM software COLMAP.
The recent progress of CNN has dramatically improved face alignment performance. However, few works have paid attention to the error-bias with respect to error distribution of facial landmarks. In this paper, we investigate the error-bias issue in face alignment, where the distributions of landmark errors tend to spread along the tangent line to landmark curves. This error-bias is not trivial since it is closely connected to the ambiguous landmark labeling task. Inspired by this observation, we seek a way to leverage the error-bias property for better convergence of CNN model. To this end, we propose anisotropic direction loss (ADL) and anisotropic attention module (AAM) for coordinate and heatmap regression, respectively. ADL imposes strong binding force in normal direction for each landmark point on facial boundaries. On the other hand, AAM is an attention module which can get anisotropic attention mask focusing on the region of point and its local edge connected by adjacent points, it has a stronger response in tangent than in normal, which means relaxed constraints in the tangent. These two methods work in a complementary manner to learn both facial structures and texture details. Finally, we integrate them into an optimized end-to-end training pipeline named ADNet. Our ADNet achieves state-of-the-art results on 300W, WFLW and COFW datasets, which demonstrates the effectiveness and robustness.
Graphs play an important role in cross-modal image-text understanding as they characterize the intrinsic structure which is robust and crucial for the measurement of cross-modal similarity. In this work, we propose a Wasserstein Coupled Graph Learning (WCGL) method to deal with the cross-modal retrieval task. First, graphs are constructed according to two input cross-modal samples separately, and passed through the corresponding graph encoders to extract robust features. Then, a Wasserstein coupled dictionary, containing multiple pairs of counterpart graph keys with each key corresponding to one modality, is constructed for further feature learning. Based on this dictionary, the input graphs can be transformed into the dictionary space to facilitate the similarity measurement through a Wasserstein Graph Embedding (WGE) process. The WGE could capture the graph correlation between the input and each corresponding key through optimal transport, and hence well characterize the inter-graph structural relationship. To further achieve discriminant graph learning, we specifically define a Wasserstein discriminant loss on the coupled graph keys to make the intra-class (counterpart) keys more compact and inter-class (non-counterpart) keys more dispersed, which further promotes the final cross-modal retrieval task. Experimental results demonstrate the effectiveness and state-of-the-art performance.
This paper proposes a framework to guide an optical flow network with external cues to achieve superior accuracy either on known or unseen domains. Given the availability of sparse yet accurate optical flow hints from an external source, these are injected to modulate the correlation scores computed by a state-of-the-art optical flow network and guide it towards more accurate predictions. Although no real sensor can provide sparse flow hints, we show how these can be obtained by combining depth measurements from active sensors with geometry and hand-crafted optical flow algorithms, leading to accurate enough hints for our purpose. Experimental results with a state-of-the-art flow network on standard benchmarks support the effectiveness of our framework, both in simulated and real conditions.
Modeling the hand-object (HO) interaction not only requires estimation of the HO pose, but also pays attention to the contact due to their interaction. Significant progress has been made in estimating hand and object separately with deep learning methods, simultaneous HO pose estimation and contact modeling has not yet been fully explored. In this paper, we present an explicit contact representation namely Contact Potential Field (CPF), and a learning-fitting hybrid framework namely MIHO to Modeling the Interaction of Hand and Object. In CPF, we treat each contacting HO vertex pair as a spring-mass system. Hence the whole system forms a potential field with minimal elastic energy at the grasp position. Extensive experiments on the two commonly used benchmarks have demonstrated that our method can achieve state-of-the-art in several reconstruction metrics, and allow us to produce more physically plausible HO pose even when the ground-truth exhibits severe interpenetration or disjointedness. Our code is available at https://github.com/lixiny/CPF.
Domain generalization (DG) aims to generalize a model trained on multiple source (i.e., training) domains to a distributionally different target (i.e., test) domain. In contrast to the DG setup that strictly requires the availability of multiple source domains, this paper considers a more realistic yet challenging scenario, namely Single Domain Generalization (SDG). In this new setting, there is only one source domain available for training, from which the limited diversity may jeopardize the model generalization on unseen target domains. To tackle this problem, we propose a style-complement module to enhance the generalization power of the model by synthesizing images from diverse distributions that are complementary to the source ones. More specifically, we adopt tractable upper and lower bounds of mutual information (MI) between the generated and source samples and perform the two-step optimization iteratively: (1) by minimizing MI upper bound approximation for each pair, the generated images are forced to diversify from the source samples; (2) subsequently, we maximize the lower bound of MI between the samples from the same semantic category, which assists the network to learn discriminative features from diverse-styled images. Extensive experiments on three benchmark datasets demonstrate the superiority of our approach, which surpasses the state-of-the-art single DG methods by up to 25.14%.
Motion is an important cue for video prediction and often utilized by separating video content into static and dynamic components. Most of the previous work utilizing motion is deterministic but there are stochastic methods that can model the inherent uncertainty of the future. Existing stochastic models either do not reason about motion explicitly or make limiting assumptions about the static part. In this paper, we reason about appearance and motion in the video stochastically by predicting the future based on the motion history. Explicit reasoning about motion without history already reaches the performance of current stochastic models. The motion history further improves the results by allowing to predict consistent dynamics several frames into the future. Our model performs comparably to the state-of-the-art models on the generic video prediction datasets, however, significantly outperforms them on two challenging real-world autonomous driving datasets with complex motion and dynamic background.
Computer vision applications such as visual relationship detection and human object interaction can be formulated as a composite (structured) set detection problem in which both the parts (subject, object, and predicate) and the sum (triplet as a whole) are to be detected in a hierarchical fashion. In this paper, we present a new approach, denoted Part-and-Sum detection Transformer (PST), to perform end-to-end visual composite set detection. Different from existing Transformers in which queries are at a single level, we simultaneously model the joint part and sum hypotheses/interactions with composite queries and attention modules. We explicitly incorporate sum queries to enable better modeling of the part-and-sum relations that are absent in the standard Transformers. Our approach also uses novel tensor-based part queries and vector-based sum queries, and models their joint interaction. We report experiments on two vision tasks, visual relationship detection and human object interaction and demonstrate that PST achieves state of the art results among single-stage models, while nearly matching the results of custom designed two-stage models.