CVPR 2021 论文列表

Convolutional neural networks have made remarkable progress in the face recognition field. The more the technology of face recognition advances, the greater discriminative features into a face template. However, this increases the threat to user privacy in case the template is exposed. In this paper, we present a modular architecture for face template protection, called IronMask, that can be combined with any face recognition system using angular distance metric. We circumvent the need for binarization, which is the main cause of performance degradation in most existing face template protections, by proposing a new real-valued error-correcting-code that is compatible with real-valued templates and can therefore, minimize performance degradation. We evaluate the efficacy of IronMask by extensive experiments on two face recognitions, ArcFace and CosFace with three datasets, CMU-Multi-PIE, FEI, and Color-FERET. According to our experimental results, IronMask achieves a true accept rate (TAR) of 99.79% at a false accept rate (FAR) of 0.0005% when combined with ArcFace, and 95.78% TAR at 0% FAR with CosFace, while providing at least 115-bit security against known attacks.

Adversarial attack arises due to the vulnerability of deep neural networks to perceive input samples injected with imperceptible perturbations. Recently, adversarial attack has been applied to visual object tracking to evaluate the robustness of deep trackers. Assuming that the model structures of deep trackers are known, a variety of white-box attack approaches to visual tracking have demonstrated promising results. However, the model knowledge about deep trackers is usually unavailable in real applications. In this paper, we propose a decision-based black-box attack method for visual object tracking. In contrast to existing black-box adversarial attack methods that deal with static images for image classification, we propose IoU attack that sequentially generates perturbations based on the predicted IoU scores from both current and historical frames. By decreasing the IoU scores, the proposed attack method degrades the accuracy of temporal coherent bounding boxes (i.e., object motions) accordingly. In addition, we transfer the learned perturbations to the next few frames to initialize temporal motion attacks. We validate the proposed IoU attack on state-of-the-art deep trackers (i.e., detection based, correlation filter based, and long-term trackers). Extensive experiments on the benchmark datasets indicate the effectiveness of the proposed IoU attack method. The source code is available at https://github.com/VISION-SJTU/IoUattack.

Convolution has been the core ingredient of modern neural networks, triggering the surge of deep learning in vision. In this work, we rethink the inherent principles of standard convolution for vision tasks, specifically spatial-agnostic and channel-specific. Instead, we present a novel atomic operation for deep neural networks by inverting the aforementioned design principles of convolution, coined as involution. We additionally demystify the recent popular self-attention operator and subsume it into our involution family as an over-complicated instantiation. The proposed involution operator could be leveraged as fundamental bricks to build the new generation of neural networks for visual recognition, powering different deep learning models on several prevalent benchmarks, including ImageNet classification, COCO detection and segmentation, together with Cityscapes segmentation. Our involution-based models improve the performance of convolutional baselines using ResNet-50 by up to 1.6% top-1 accuracy, 2.5% and 2.4% bounding box AP, and 4.7% mean IoU absolutely while compressing the computational cost to 66%, 65%, 72%, and 57% on the above benchmarks, respectively. Code and pre-trained models for all the tasks are available at https://github.com/d-li14/involution.

Physical adversarial examples for camera-based computer vision have so far been achieved through visible artifacts -- a sticker on a Stop sign, colorful borders around eyeglasses or a 3D printed object with a colorful texture. An implicit assumption here is that the perturbations must be visible so that a camera can sense them. By contrast, we contribute a procedure to generate, for the first time, physical adversarial examples that are invisible to human eyes. Rather than modifying the victim object with visible artifacts, we modify light that illuminates the object. We demonstrate how an attacker can craft a modulated light signal that adversarially illuminates a scene and causes targeted misclassifications on a state-of-the-art ImageNet deep learning model. Concretely, we exploit the radiometric rolling shutter effect in commodity cameras to create precise striping patterns that appear on images. To human eyes, it appears like the object is illuminated, but the camera creates an image with stripes that will cause ML models to output the attacker-desired classification. We conduct a range of simulation and physical experiments with LEDs, demonstrating targeted attack rates up to 84%.

Given a monocular face image as input, 3D face geometry reconstruction aims to recover a corresponding 3Dface mesh. Recently, both optimization-based and learning-based face reconstruction methods have taken advantage of the emerging differentiable renderer and shown promising results. However, the differentiable renderer, mainly based on graphics rules, simplifies the realistic mechanism of the illumination, reflection, etc., of the real world, thus can-not produce realistic images. This brings a lot of domain-shift noise to the optimization or training process. In this work, we introduce a novel Generative Adversarial Renderer (GAR) and propose to tailor its inverted version to the general fitting pipeline, to tackle the above problem. Specifically, the carefully designed neural renderer takes a face normal map and a latent code representing other fac-tors as inputs and renders a realistic face image. Since the GAR learns to model the complicated real-world image, instead of relying on the simplified graphics rules, it is capable of producing realistic images, which essentially inhibits the domain-shift noise in training and optimization. Equipped with the elaborated GAR, we further proposed a novel approach to predict 3D face parameters, in which we first obtain fine initial parameters via Renderer Invertingand then refine it with gradient-based optimizers. Extensive experiments have been conducted to demonstrate the effectiveness of the proposed generative adversarial renderer and the novel optimization-based face reconstruction framework. Our method achieves state-of-the-art performance on multiple face reconstruction datasets.

Unprocessed RAW data is a highly valuable image format for image editing and computer vision. However, since the file size of RAW data is huge, most users can only get access to processed and compressed sRGB images. To bridge this gap, we design an Invertible Image Signal Processing (InvISP) pipeline, which not only enables rendering visually appealing sRGB images but also allows recovering nearly perfect RAW data. Due to our framework's inherent reversibility, we can reconstruct realistic RAW data instead of synthesizing RAW data from sRGB images without any memory overhead. We also integrate a differentiable JPEG compression simulator that empowers our framework to reconstruct RAW data from JPEG images. Extensive quantitative and qualitative experiments on two DSLR demonstrate that our method obtains much higher quality in both rendered sRGB images and reconstructed RAW data than alternative methods.

Invertible networks have various benefits for image denoising since they are lightweight, information-lossless, and memory-saving during back-propagation. However, applying invertible models to remove noise is challenging because the input is noisy, and the reversed output is clean, following two different distributions. We propose an invertible denoising network, InvDN, to address this challenge. InvDN transforms the noisy input into a low-resolution clean image and a latent representation containing noise. To discard noise and restore the clean image, InvDN replaces the noisy latent representation with another one sampled from a prior distribution during reversion. The denoising performance of InvDN is better than all the existing competitive models, achieving a new state-of-the-art result for the SIDD dataset while enjoying less run time. Moreover, the size of InvDN is far smaller, only having 4.2% of the number of parameters compared to the most recently proposed DANet. Further, via manipulating the noisy latent representation, InvDN is also able to generate noise more similar to the original one. Our code is available at: https://github.com/Yang-Liu1082/InvDN.git.

We present a novel boundary-aware loss term for semantic segmentation using an inverse-transformation network, which efficiently learns the degree of parametric transformations between estimated and target boundaries. This plug-in loss term complements the cross-entropy loss in capturing boundary transformations and allows consistent and significant performance improvement on segmentation backbone models without increasing their size and computational complexity. We analyze the quantitative and qualitative effects of our loss function on three indoor and outdoor segmentation benchmarks, including Cityscapes, NYU-Depth-v2, and PASCAL, integrating it into the training phase of several backbone networks in both single-task and multi-task settings. Our extensive experiments show that the proposed method consistently outperforms baselines, and even sets the new state-of-the-art on two datasets.

This paper studies the problem of inverse cloth simulation---to estimate shape and time-varying poses of the underlying body that generates physically plausible cloth motion, which matches to the point cloud measurements on the clothed humans. A key innovation is to represent the dynamics of the cloth geometry using a dynamical system that is controlled by the body states (shape and pose). This allows us to express the cloth motion as a resultant of external (skin friction and gravity) and internal (elasticity) forces. Inspired by the theory of optimal control, we optimize the body states such that the simulated cloth motion is matched to the point cloud measurements, and the analytic gradient of the simulator is back-propagated to update the body states. We propose a cloth relaxation scheme to initialize the cloth state, which ensures the physical validity. Our method produces physically plausible and temporally smooth cloth and body movements that are faithful to the measurements, and shows superior performance compared to the existing methods. As a byproduct, the stress and strain that are applied to the body and clothes can be recovered.

Predicting human trajectories is an important component of autonomous moving platforms, such as social robots and self-driving cars. Human trajectories are affected by both the physical features of the environment and social interactions with other humans. Despite recent surge of studies on human path prediction, most works focus on static scene information, therefore, cannot leverage the rich dynamic visual information of the scene. In this work, we propose Introvert, a model which predicts human path based on his/her observed trajectory and the dynamic scene context, captured via a conditional 3D visual attention mechanism working on the input video. Introvert infers both environment constraints and social interactions through observing the dynamic scene instead of communicating with other humans, hence, its computational cost is independent of how crowded the surrounding of a target human is. In addition, to focus on relevant interactions and constraints for each human, Introvert conditions its 3D attention model on the observed trajectory of the target human to extract and focus on relevant spatio-temporal primitives. Our experiments on five publicly available datasets show that the Introvert improves the prediction errors of the state of the art.

Compositing an image usually inevitably suffers from inharmony problem that is mainly caused by incompatibility of foreground and background from two different images with distinct surfaces and lights, corresponding to material-dependent and light-dependent characteristics, namely, reflectance and illumination intrinsic images, respectively. Therefore, we seek to solve image harmonization via separable harmonization of reflectance and illumination, i.e., intrinsic image harmonization. Our method is based on an autoencoder that disentangles composite image into reflectance and illumination for further separate harmonization. Specifically, we harmonize reflectance through material-consistency penalty, while harmonize illumination by learning and transferring light from background to foreground, moreover, we model patch relations between foreground and background of composite images in an inharmony-free learning way, to adaptively guide our intrinsic image harmonization. Both extensive experiments and ablation studies demonstrate the power of our method as well as the efficacy of each component. We also contribute a new challenging dataset for benchmarking illumination harmonization. Code and dataset are at https://github.com/zhenglab/IntrinsicHarmony.

Most of unsupervised person Re-Identification (Re-ID) works produce pseudo-labels by measuring the feature similarity without considering the distribution discrepancy among cameras, leading to degraded accuracy in label computation across cameras. This paper targets to address this challenge by studying a novel intra-inter camera similarity for pseudo-label generation. We decompose the sample similarity computation into two stage, i.e., the intra-camera and inter-camera computations, respectively. The intra-camera computation directly leverages the CNN features for similarity computation within each camera. Pseudo-labels generated on different cameras train the re-id model in a multi-branch network. The second stage considers the classification scores of each sample on different cameras as a new feature vector. This new feature effectively alleviates the distribution discrepancy among cameras and generates more reliable pseudo-labels. We hence train our re-id model in two stages with intra-camera and inter-camera pseudo-labels, respectively. This simple intra-inter camera similarity produces surprisingly good performance on multiple datasets, e.g., achieves rank-1 accuracy of 89.5% on the Market1501 dataset, outperforming the recent unsupervised works by 9+%, and is comparable with the latest transfer learning works that leverage extra annotations.

Video grounding aims to localize a moment from an untrimmed video for a given textual query. Existing approaches focus more on the alignment of visual and language stimuli with various likelihood-based matching or regression strategies, i.e., P(Y|X). Consequently, these models may suffer from spurious correlations between the language and video features due to the selection bias of the dataset. 1) To uncover the causality behind the model and data, we first propose a novel paradigm from the perspective of the causal inference, i.e., interventional video grounding (IVG) that leverages backdoor adjustment to deconfound the selection bias based on structured causal model (SCM) and do-calculus P(Y|do(X)). Then, we present a simple yet effective method to approximate the unobserved confounder as it cannot be directly sampled from the dataset. 2) Meanwhile, we introduce a dual contrastive learning approach (DCL) to better align the text and video by maximizing the mutual information (MI) between query and video clips, and the MI between start/end frames of a target moment and the others within a video to learn more informative visual representations. Experiments on three standard benchmarks show the effectiveness of our approaches.

Image super-resolution (SR) techniques have been developing rapidly, benefiting from the invention of deep networks and its successive breakthroughs. However, it is acknowledged that deep learning and deep neural networks are difficult to interpret. SR networks inherit this mysterious nature and little works make attempt to understand them. In this paper, we perform attribution analysis of SR networks, which aims at finding the input pixels that strongly influence the SR results. We propose a novel attribution approach called local attribution map (LAM), which inherits the integral gradient method yet with two unique features. One is to use the blurred image as the baseline input, and the other is to adopt the progressive blurring function as the path function. Based on LAM, we show that: (1) SR networks with a wider range of involved input pixels could achieve better performance. (2) Attention networks and non-local networks extract features from a wider range of input pixels. (3) Comparing with the range that actually contributes, the receptive field is large enough for most deep networks. (4) For SR networks, textures with regular stripes or grids are more likely to be noticed, while complex semantics are difficult to utilize. Our work opens new directions for designing SR networks and interpreting low-level vision deep models.

Human trajectory forecasting in crowds, at its core, is a sequence prediction problem with specific challenges of capturing inter-sequence dependencies (social interactions) and consequently predicting socially-compliant multimodal distributions. In recent years, neural network-based methods have been shown to outperform hand-crafted methods on distance-based metrics. However, these data-driven methods still suffer from one crucial limitation: lack of interpretability. To overcome this limitation, we leverage the power of discrete choice models to learn interpretable rule-based intents, and subsequently utilise the expressibility of neural networks to model scene-specific residual. Extensive experimentation on the interaction-centric benchmark TrajNet++ demonstrates the effectiveness of our proposed architecture to explain its predictions without compromising the accuracy.

Despite the data labeling cost for the object detection tasks being substantially more than that of the classification tasks, semi-supervised learning methods for object detection have not been studied much. In this paper, we propose an Interpolation-based Semi-supervised learning method for object Detection (ISD), which considers and solves the problems caused by applying conventional Interpolation Regularization (IR) directly to object detection. We divide the output of the model into two types according to the objectness scores of both original patches that are mixed in IR. Then, we apply a separate loss suitable for each type in an unsupervised manner. The proposed losses dramatically improve the performance of semi-supervised learning as well as supervised learning. In the supervised learning setting, our method improves the baseline methods by a significant margin. In the semi-supervised learning setting, our algorithm improves the performance on a benchmark dataset (PASCAL VOC and MSCOCO) in a benchmark architecture (SSD).

The goal of semi-supervised object detection is to learn a detection model using only a few labeled data and large amounts of unlabeled data, thereby reducing the cost of data labeling. Although a few studies have proposed various self-training-based methods or consistency regularization-based methods, they ignore the discrepancies among the detection results in the same image that occur during different training iterations. Additionally, the predicted detection results vary among different detection models. In this paper, we propose an interactive form of self-training using mean teachers for semi-supervised object detection. Specifically, to alleviate the instability among the detection results in different iterations, we propose using nonmaximum suppression to fuse the detection results from different iterations. Simultaneously, we use multiple detection heads that predict pseudo labels for each other to provide complementary information. Furthermore, to avoid different detection heads collapsing to each other, we use a mean teacher model instead of the original detection model to predict the pseudo labels. Thus, the object detection model can be trained on both labeled and unlabeled data. Extensive experimental results verify the effectiveness of our proposed method.

An image is worth a thousand words, conveying information that goes beyond the physical visual content therein. In this paper, we study the intent behind social media images with an aim to analyze how visual information can help the recognition of human intent. Towards this goal, we introduce an intent dataset, Intentonomy, comprising 14K images covering a wide range of everyday scenes. These images are manually annotated with 28 intent categories that are derived from a social psychology taxonomy. We then systematically study whether, and to what extent, commonly used visual information, i.e., object and context, contribute to human motive understanding. Based on our findings, we conduct further study to quantify the effect of attending to object and context classes as well as textual information in the form of hashtags when training an intent classifier. Our results quantitatively and qualitatively shed light on how visual and textual information can produce observable effects when predicting intent.

Daily human activities, e.g., locomotion, exercises, and resting, are heavily guided by the tactile interactions between the human and the ground. In this work, leveraging such tactile interactions, we propose a 3D human pose estimation approach using the pressure maps recorded by a tactile carpet as input. We build a low-cost, high-density, large-scale intelligent carpet, which enables the real-time recordings of human-floor tactile interactions in a seamless manner. We collect a synchronized tactile and visual dataset on various human activities. Employing a state-of-the-art camera-based pose estimation model as supervision, we design and implement a deep neural network model to infer 3D human poses using only the tactile information. Our pipeline can be further scaled up to multi-person pose estimation. We evaluate our system and demonstrate its potential applications in diverse fields.