CVPR 2019 论文列表

Image compression-based approaches for defending against the adversarial-example attacks, which threaten the safety use of deep neural networks (DNN), have been investigated recently. However, prior works mainly rely on directly tuning parameters like compression rate, to blindly reduce image features, thereby lacking guarantee on both defense efficiency (i.e. accuracy of polluted images) and classification accuracy of benign images, after applying defense methods. To overcome these limitations, we propose a JPEG-based defensive compression framework, namely "feature distillation", to effectively rectify adversarial examples without impacting classification accuracy on benign data. Our framework significantly escalates the defense efficiency with marginal accuracy reduction using a twostep method: First, we maximize malicious features filtering of adversarial input perturbations by developing defensive quantization in frequency domain of JPEG compression or decompression, guided by a semi-analytical method; Second, we suppress the distortions of benign features to restore classification accuracy through a DNN-oriented quantization refine process. Our experimental results show that proposed "feature distillation" can significantly surpass the latest input-transformation based mitigations such as Quilting and TV Minimization in three aspects, including defense efficiency (improve classification accuracy from 20% to 90% on adversarial examples), accuracy of benign images after defense (<= 1% accuracy degradation), and processing time per image ( 259x Speedup). Moreover, our solution also can provide the best defense efficiency ( 60% accuracy) against the latest BPDA attack with least accuracy reduction ( 1%) on benign images among all other input-transformation based defense methods.

Adversarial attacks to image classification systems present challenges to convolutional networks and opportunities for understanding them. This study suggests that adversarial perturbations on images lead to noise in the features constructed by these networks. Motivated by this observation, we develop new network architectures that increase adversarial robustness by performing feature denoising. Specifically, our networks contain blocks that denoise the features using non-local means or other filters; the entire networks are trained end-to-end. When combined with adversarial training, our feature denoising networks substantially improve the state-of-the-art in adversarial robustness in both white-box and black-box attack settings. On ImageNet, under 10-iteration PGD white-box attacks where prior art has 27.9% accuracy, our method achieves 55.7%; even under extreme 2000-iteration PGD white-box attacks, our method secures 42.6% accuracy. Our method was ranked first in Competition on Adversarial Attacks and Defenses (CAAD) 2018 --- it achieved 50.6% classification accuracy on a secret, ImageNet-like test dataset against 48 unknown attackers, surpassing the runner-up approach by 10%. Code is available at https://github.com/facebookresearch/ImageNet-Adversarial-Training.

The search for predictive models that generalize to the long tail of sensor inputs is the central difficulty when developing data-driven models for autonomous vehicles. In this paper, we use lane detection to study modeling and training techniques that yield better performance on real world test drives. On the modeling side, we introduce a novel fully convolutional model of lane detection that learns to decode lane structures instead of delegating structure inference to post-processing. In contrast to previous works, our convolutional decoder is able to represent an arbitrary number of lanes per image, preserves the polyline representation of lanes without reducing lanes to polynomials, and draws lanes iteratively without requiring the computational and temporal complexity of recurrent neural networks. Because our model includes an estimate of the joint distribution of neighboring pixels belonging to the same lane, our formulation includes a natural and computationally cheap definition of uncertainty. On the training side, we demonstrate a simple yet effective approach to adapt the model to new environments using unsupervised style transfer. By training FastDraw to make predictions of lane structure that are invariant to low-level stylistic differences between images, we achieve strong performance at test time in weather and lighting conditions that deviate substantially from those of the annotated datasets that are publicly available. We quantitatively evaluate our approach on the CVPR 2017 Tusimple lane marking challenge, difficult CULane datasets [29], and a small labeled dataset of our own and achieve competitive accuracy while running at 90 FPS.

Image captioning is an ambiguous problem, with many suitable captions for an image. To address ambiguity, beam search is the de facto method for sampling multiple captions. However, beam search is computationally expensive and known to produce generic captions. To address this concern, some variational auto-encoder (VAE) and generative adversarial net (GAN) based methods have been proposed. Though diverse, GAN and VAE are less accurate. In this paper, we first predict a meaningful summary of the image, then generate the caption based on that summary. We use part-of-speech as summaries, since our summary should drive caption generation. We achieve the trifecta: (1) High accuracy for the diverse captions as evaluated by standard captioning metrics and user studies; (2) Faster computation of diverse captions compared to beam search and diverse beam search; and (3) High diversity as evaluated by counting novel sentences, distinct n-grams and mutual overlap (i.e., mBleu-4) scores.

This paper addresses the problem of 3D pose estimation for multiple people in a few calibrated camera views. The main challenge of this problem is to find the cross-view correspondences among noisy and incomplete 2D pose predictions. Most previous methods address this challenge by directly reasoning in 3D using a pictorial structure model, which is inefficient due to the huge state space. We propose a fast and robust approach to solve this problem. Our key idea is to use a multi-way matching algorithm to cluster the detected 2D poses in all views. Each resulting cluster encodes 2D poses of the same person across different views and consistent correspondences across the keypoints, from which the 3D pose of each person can be effectively inferred. The proposed convex optimization based multi-way matching algorithm is efficient and robust against missing and false detections, without knowing the number of people in the scene. Moreover, we propose to combine geometric and appearance cues for cross-view matching. The proposed approach achieves significant performance gains from the state-of-the-art (96.3% vs. 90.6% and 96.9% vs. 88% on the Campus and Shelf datasets, respectively), while being efficient for real-time applications.

We present a new, fast and flexible pipeline for indoor scene synthesis that is based on deep convolutional generative models. Our method operates on a top-down image-based representation, and inserts objects iteratively into the scene by predict their category, location, orientation and size with separate neural network modules. Our pipeline naturally supports automatic completion of partial scenes, as well as synthesis of complete scenes, without any modifications. Our method is significantly faster than the previous image-based method, and generates results that outperforms it and other state-of-the-art deep generative scene models in terms of faithfulness to training data and perceived visual quality.

We present a deep learning method for the interactive video object segmentation. Our method is built upon two core operations, interaction and propagation, and each operation is conducted by Convolutional Neural Networks. The two networks are connected both internally and externally so that the networks are trained jointly and interact with each other to solve the complex video object segmentation problem. We propose a new multi-round training scheme for the interactive video object segmentation so that the networks can learn how to understand the user's intention and update incorrect estimations during the training. At the testing time, our method produces high-quality results and also runs fast enough to work with users interactively. We evaluated the proposed method quantitatively on the interactive track benchmark at the DAVIS Challenge 2018. We outperformed other competing methods by a significant margin in both the speed and the accuracy. We also demonstrated that our method works well with real user interactions.

Recently, deep learning based video super-resolution (SR) methods have achieved promising performance. To simultaneously exploit the spatial and temporal information of videos, employing 3-dimensional (3D) convolutions is a natural approach. However, straight utilizing 3D convolutions may lead to an excessively high computational complexity which restricts the depth of video SR models and thus undermine the performance. In this paper, we present a novel fast spatio-temporal residual network (FSTRN) to adopt 3D convolutions for the video SR task in order to enhance the performance while maintaining a low computational load. Specifically, we propose a fast spatio-temporal residual block (FRB) that divide each 3D filter to the product of two 3D filters, which have considerably lower dimensions. Furthermore, we design a cross-space residual learning that directly links the low-resolution space and the high-resolution space, which can greatly relieve the computational burden on the feature fusion and up-scaling parts. Extensive evaluations and comparisons on benchmark datasets validate the strengths of the proposed approach and demonstrate that the proposed network significantly outperforms the current state-of-the-art methods.

We propose a real-time method to estimate spatially-varying indoor lighting from a single RGB image. Given an image and a 2D location in that image, our CNN estimates a 5th order spherical harmonic representation of the lighting at the given location in less than 20ms on a laptop mobile graphics card. While existing approaches estimate a single, global lighting representation or require depth as input, our method reasons about local lighting without requiring any geometry information. We demonstrate, through quantitative experiments including a user study, that our results achieve lower lighting estimation errors and are preferred by users over the state-of-the-art. Our approach can be used directly for augmented reality applications, where a virtual object is relit realistically at any position in the scene in real-time.

Removing undesired reflections from images taken through the glass is of great importance in computer vision. It serves as a means to enhance the image quality for aesthetic purposes as well as to preprocess images in machine learning and pattern recognition applications. We propose a convex model to suppress the reflection from a single input image. Our model implies a partial differential equation with gradient thresholding, which is solved efficiently using Discrete Cosine Transform. Extensive experiments on synthetic and real-world images demonstrate that our approach achieves desirable reflection suppression results and dramatically reduces the execution time compared to the state of the art.

In this paper we illustrate how to perform both visual object tracking and semi-supervised video object segmentation, in real-time, with a single simple approach. Our method, dubbed SiamMask, improves the offline training procedure of popular fully-convolutional Siamese approaches for object tracking by augmenting their loss with a binary segmentation task. Once trained, SiamMask solely relies on a single bounding box initialisation and operates online, producing class-agnostic object segmentation masks and rotated bounding boxes at 55 frames per second. Despite its simplicity, versatility and fast speed, our strategy allows us to establish a new state-of-the-art among real-time trackers on VOT-2018, while at the same time demonstrating competitive performance and the best speed for the semi-supervised video object segmentation task on DAVIS-2016 and DAVIS-2017.

Object class labelling is the task of annotating images with labels on the presence or absence of objects from a given class vocabulary. Simply asking one yes-no question per class, however, has a cost that is linear in the vocabulary size and is thus inefficient for large vocabularies. Modern approaches rely on a hierarchical organization of the vocabulary to reduce annotation time, but remain expensive (several minutes per image for the 200 classes in ILSVRC). Instead, we propose a new interface where classes are annotated via speech. Speaking is fast and allows for direct access to the class name, without searching through a list or hierarchy. As additional advantages, annotators can simultaneously speak and scan the image for objects, the interface can be kept extremely simple, and using it requires less mouse movement. As annotators using our interface should only say words from a given class vocabulary, we propose a dedicated task to train them to do so. Through experiments on COCO and ILSVRC, we show our method yields high-quality annotations at 2.3x -14.9x less annotation time than existing methods.

Automated design of neural network architectures tailored for a specific task is an extremely promising, albeit inherently difficult, avenue to explore. While most results in this domain have been achieved on image classification and language modelling problems, here we concentrate on dense per-pixel tasks, in particular, semantic image segmentation using fully convolutional networks. In contrast to the aforementioned areas, the design choices of a fully convolutional network require several changes, ranging from the sort of operations that need to be used - e.g., dilated convolutions - to a solving of a more difficult optimisation problem. In this work, we are particularly interested in searching for high-performance compact segmentation architectures, able to run in real-time using limited resources. To achieve that, we intentionally over-parameterise the architecture during the training time via a set of auxiliary cells that provide an intermediate supervisory signal and can be omitted during the evaluation phase. The design of the auxiliary cell is emitted by a controller, a neural network with the fixed structure trained using reinforcement learning. More crucially, we demonstrate how to efficiently search for these architectures within limited time and computational budgets. In particular, we rely on a progressive strategy that terminates non-promising architectures from being further trained, and on Polyak averaging coupled with knowledge distillation to speed-up the convergence. Quantitatively, in 8 GPU-days our approach discovers a set of architectures performing on-par with state-of-the-art among compact models on the semantic segmentation, pose estimation and depth prediction tasks. Code will be made available here: https://github.com/drsleep/nas-segm-pytorch

Manually labeling objects by tracing their boundaries is a laborious process. In Polygon-RNN++, the authors proposed Polygon-RNN that produces polygonal annotations in a recurrent manner using a CNN-RNN architecture, allowing interactive correction via humans-in-the-loop. We propose a new framework that alleviates the sequential nature of Polygon-RNN, by predicting all vertices simultaneously using a Graph Convolutional Network (GCN). Our model is trained end-to-end, and runs in real time. It supports object annotation by either polygons or splines, facilitating labeling efficiency for both line-based and curved objects. We show that Curve-GCN outperforms all existing approaches in automatic mode, including the powerful DeepLab, and is significantly more efficient in interactive mode than Polygon-RNN++. Our model runs at 29.3ms in automatic, and 2.6ms in interactive mode, making it 10x and 100x faster than Polygon-RNN++.

Existing human pose estimation approaches often only consider how to improve the model generalisation performance, but putting aside the significant efficiency problem. This leads to the development of heavy models with poor scalability and cost-effectiveness in practical use. In this work, we investigate the under-studied but practically critical pose model efficiency problem. To this end, we present a new Fast Pose Distillation (FPD) model learning strategy. Specifically, the FPD trains a lightweight pose neural network architecture capable of executing rapidly with low computational cost. It is achieved by effectively transferring the pose structure knowledge of a strong teacher network. Extensive evaluations demonstrate the advantages of our FPD method over a broad range of state-of-the-art pose estimation approaches in terms of model cost-effectiveness on two standard benchmark datasets, MPII Human Pose and Leeds Sports Pose.

Dialog is an effective way to exchange information, but subtle details and nuances are extremely important. While significant progress has paved a path to address visual dialog with algorithms, details and nuances remain a challenge. Attention mechanisms have demonstrated compelling results to extract details in visual question answering and also provide a convincing framework for visual dialog due to their interpretability and effectiveness. However, the many data utilities that accompany visual dialog challenge existing attention techniques. We address this issue and develop a general attention mechanism for visual dialog which operates on any number of data utilities. To this end, we design a factor graph based attention mechanism which combines any number of utility representations. We illustrate the applicability of the proposed approach on the challenging and recently introduced VisDial datasets, outperforming recent state-of-the-art methods by 1.1% for VisDial0.9 and by 2% for VisDial1.0 on MRR. Our ensemble model improved the MRR score on VisDial1.0 by more than 6%.

Emotion recognition from facial expressions is an interesting and challenging problem and has attracted much attention in recent years. Substantial previous research has only been able to address the ambiguity of "what describes the expression", which assumes that each facial expression is associated with one or more predefined affective labels while ignoring the fact that multiple emotions always have different intensities in a single picture. Therefore, to depict facial expressions more accurately, this paper adopts a label distribution learning approach for emotion recognition that can address the ambiguity of "how to describe the expression" and proposes an emotion distribution learning method that exploits label correlations locally. Moreover, a local low-rank structure is employed to capture the local label correlations implicitly. Experiments on benchmark facial expression datasets demonstrate that our method can better address the emotion distribution recognition problem than state-of-the-art methods.

Automated deception detection (ADD) from real-life videos is a challenging task. It specifically needs to address two problems: (1) Both face and body contain useful cues regarding whether a subject is deceptive. How to effectively fuse the two is thus key to the effectiveness of an ADD model. (2) Real-life deceptive samples are hard to collect; learning with limited training data thus challenges most deep learning based ADD models. In this work, both problems are addressed. Specifically, for face-body multimodal learning, a novel face-focused cross-stream network (FFCSN) is proposed. It differs significantly from the popular two-stream networks in that: (a) face detection is added into the spatial stream to capture the facial expressions explicitly, and (b) correlation learning is performed across the spatial and temporal streams for joint deep feature learning across both face and body. To address the training data scarcity problem, our FFCSN model is trained with both meta learning and adversarial learning. Extensive experiments show that our FFCSN model achieves state-of-the-art results. Further, the proposed FFCSN model as well as its robust training strategy are shown to be generally applicable to other human-centric video analysis tasks such as emotion recognition from user-generated videos.

Face parsing computes pixel-wise label maps for different semantic components (e.g., hair, mouth, eyes) from face images. Existing face parsing literature have illustrated significant advantages by focusing on individual regions of interest (RoIs) for faces and facial components. However,the traditional crop-and-resize focusing mechanism ignores all contextual area outside the RoIs, and thus is not suitable when the component area is unpredictable, e.g. hair. Inspired by the physiological vision system of human, we propose a novel RoI Tanh-warping operator that combines the central vision and the peripheral vision together. It addresses the dilemma between a limited sized RoI for focusing and an unpredictable area of surrounding context for peripheral information. To this end, we propose a novel hybrid convolutional neural network for face parsing. It uses hierarchical local based method for inner facial components and global methods for outer facial components. The whole framework is simple and principled, and can be trained end-to-end. To facilitate future research of face parsing, we also manually relabel the training data of the HELEN dataset and will make it public. Experiments on both HELEN and LFW-PL benchmarks demonstrate that our method surpasses state-of-the-art methods.