ICCV 2017 论文列表

Since convolutional neural network (CNN) lacks an inherent mechanism to handle large scale variations, we always need to compute feature maps multiple times for multi-scale object detection, which has the bottleneck of computational cost in practice. To address this, we devise a recurrent scale approximation (RSA) to compute feature map once only, and only through this map can we approximate the rest maps on other levels. At the core of RSA is the recursive rolling out mechanism: given an initial map on a particular scale, it generates the prediction on a smaller scale that is half the size of input. To further increase efficiency and accuracy, we (a): design a scale-forecast network to globally predict potential scales in the image since there is no need to compute maps on all levels of the pyramid. (b): propose a landmark retracing network (LRN) to retrace back locations of the regressed landmarks and generate a confidence score for each landmark; LRN can effectively alleviate false positives due to the accumulated error in RSA. The whole system could be trained end-to-end in a unified CNN framework. Experiments demonstrate that our proposed algorithm is superior against state-of-the-arts on face detection benchmarks and achieves comparable results for generic proposal generation. The source code of RSA is available at github.com/sciencefans/RSA-for-object-detection.

In this paper we are interested in the problem of image segmentation given natural language descriptions, i.e. referring expressions. Existing works tackle this problem by first modeling images and sentences independently and then segment images by combining these two types of representations. We argue that learning word-to-image interaction is more native in the sense of jointly modeling two modalities for the image segmentation task, and we propose convolutional multimodal LSTM to encode the sequential interactions between individual words, visual information, and spatial information. We show that our proposed model outperforms the baseline model on benchmark datasets. In addition, we analyze the intermediate output of the proposed multimodal LSTM approach and empirically explain how this approach enforces a more effective word-to-image interaction.

This work proposes Recurrent Neural Network (RNN) models to predict structured 'image situations' -- actions and noun entities fulfilling semantic roles related to the action. In contrast to prior work relying on Conditional Random Fields (CRFs), we use a specialized action prediction network followed by an RNN for noun prediction. Our system obtains state-of-the-art accuracy on the challenging recent imSitu dataset, beating CRF-based models, including ones trained with additional data. Further, we show that specialized features learned from situation prediction can be transferred to the task of image captioning to more accurately describe human-object interactions.

We introduce a novel formulation of temporal color constancy which considers multiple frames preceding the frame for which illumination is estimated. We propose an end-to-end trainable recurrent color constancy network -- the RCC-Net -- which exploits convolutional LSTMs and a simulated sequence to learn compositional representations in space and time. We use a standard single frame color constancy benchmark, the SFU Gray Ball Dataset, which can be adapted to a temporal setting. Extensive experiments show that the proposed method consistently outperforms single-frame state-of-the-art methods and their temporal variants.

Despite remarkable progress of face analysis techniques, detecting landmarks on large-pose faces is still difficult due to self-occlusion, subtle landmark difference and incomplete information. To address these challenging issues, we introduce a novel recurrent 3D-2D dual learning model that alternatively performs 2D-based 3D face model refinement and 3D-to-2D projection based 2D landmark refinement to reliably reason about self-occluded landmarks, precisely capture the subtle landmark displacement and accurately detect landmarks even in presence of extremely large poses. The proposed model presents the first loop-closed learning framework that effectively exploits the informative feedback from the 3D-2D learning and its dual 2D-3D refinement tasks in a recurrent manner. Benefiting from these two mutual-boosting steps, our proposed model demonstrates appealing robustness to large poses (up to profile pose) and outstanding ability to capture fine-scale landmark displacement compared with existing 3D models. It achieves new state-of-the-art on the challenging AFLW benchmark. Moreover, our proposed model introduces a new architectural design that economically utilizes intermediate features and achieves 4x faster speed than its deep learning based counterparts.

Deep neural networks (DNNs) trained on large-scale datasets have recently achieved impressive improvements in face recognition. But a persistent challenge remains to develop methods capable of handling large pose variations that are relatively under-represented in training data. This paper presents a method for learning a feature representation that is invariant to pose, without requiring extensive pose coverage in training data. We first propose to generate non-frontal views from a single frontal face, in order to increase the diversity of training data while preserving accurate facial details that are critical for identity discrimination. Our next contribution is to seek a rich embedding that encodes identity features, as well as non-identity ones such as pose and landmark locations. Finally, we propose a new feature reconstruction metric learning to explicitly disentangle identity and pose, by demanding alignment between the feature reconstructions through various combinations of identity and pose features, which is obtained from two images of the same subject. Experiments on both controlled and in-the-wild face datasets, such as MultiPIE, 300WLP and the profile view database CFP, show that our method consistently outperforms the state-of-the-art, especially on images with large head pose variations.

Advancements in deep learning have ignited an explosion of research on efficient hardware for embedded computer vision. Hardware vision acceleration, however, does not address the cost of capturing and processing the image data that feeds these algorithms. We examine the role of the image signal processing (ISP) pipeline in computer vision to identify opportunities to reduce computation and save energy. The key insight is that imaging pipelines should be designed to be configurable: to switch between a traditional photography mode and a low-power vision mode that produces lower-quality image data suitable only for computer vision. We use eight computer vision algorithms and a reversible pipeline simulation tool to study the imaging system's impact on vision performance. For both CNN-based and classical vision algorithms, we observe that only two ISP stages, demosaicing and gamma compression, are critical for task performance. We propose a new image sensor design that can compensate for skipping these stages. The sensor design features an adjustable resolution and tunable analog-to-digital converters (ADCs). Our proposed imaging system's vision mode disables the ISP entirely and configures the sensor to produce subsampled, lower-precision image data. This vision mode can save 75% of the average energy of a baseline photography mode while having only a small impact on vision task accuracy.

Most previous algorithms for the recognition of Action Units (AUs) were trained on a small number of sample images. This was due to the limited amount of labeled data available at the time. This meant that data-hungry deep neural networks, which have shown their potential in other computer vision problems, could not be successfully trained to detect AUs. A recent publicly available database with close to a million labeled images has made this training possible. Image and individual variability (e.g., pose, scale, illumination, ethnicity) in this set is very large. Unfortunately, the labels in this dataset are not perfect (i.e., they are noisy), making convergence of deep nets difficult. To harness the richness of this dataset while being robust to the inaccuracies of the labels, we derive a novel global-local loss. This new loss function is shown to yield fast globally meaningful convergences and locally accurate results. Comparative results with those of the EmotioNet challenge demonstrate that our newly derived loss yields superior recognition of AUs than state-of-the-art algorithms.

We present a framework for learning to describe fine-grained visual differences between instances using attribute phrases. Attribute phrases capture distinguishing aspects of an object (e.g., "propeller on the nose" or "door near the wing" for airplanes) in a compositional manner. Instances within a category can be described by a set of these phrases and collectively they span the space of semantic attributes for a category. We collect a large dataset of such phrases by asking annotators to describe several visual differences between a pair of instances within a category. We then learn to describe and ground these phrases to images in the context of a *reference game* between a speaker and a listener. The goal of a speaker is to describe attributes of an image that allows the listener to correctly identify it within a pair. Data collected in a pairwise manner improves the ability of the speaker to generate, and the ability of the listener to interpret visual descriptions. Moreover, due to the compositionality of attribute phrases, the trained listeners can interpret descriptions not seen during training for image retrieval, and the speakers can generate attribute-based explanations for differences between previously unseen categories. We also show that embedding an image into the semantic space of attribute phrases derived from listeners offers 20% improvement in accuracy over existing attribute-based representations on the FGVC-aircraft dataset.

We present a novel method to realistically puppeteer and animate a face from a single RGB image using a source video sequence. We begin by fitting a multilinear PCA model to obtain the 3D geometry and a single texture of the target face. In order for the animation to be realistic, however, we need dynamic per-frame textures that capture subtle wrinkles and deformations corresponding to the animated facial expressions. This problem is highly underconstrained, as dynamic textures cannot be obtained directly from a single image. Furthermore, if the target face has a closed mouth, it is not possible to obtain actual images of the mouth interior. To address this issue, we train a Deep Generative Network that can infer realistic per-frame texture deformations, including the mouth interior, of the target identity using the per-frame source textures and the single target texture. By retargeting the PCA expression geometry from the source, as well as using the newly inferred texture, we can both animate the face and perform video face replacement on the source video using the target appearance.

We present a novel approach to 6DOF pose estimation and segmentation of rigid 3D objects using a single monocular RGB camera based on temporally consistent, local color histograms. We show that this approach outperforms previous methods in cases of cluttered backgrounds, heterogenous objects, and occlusions. The proposed histograms can be used as statistical object descriptors within a template matching strategy for pose recovery after temporary tracking loss e. g. caused by massive occlusion or if the object leaves the camera's field of view. The descriptors can be trained online within a couple of seconds moving a handheld object in front of a camera. During the training stage, our approach is already capable to recover from accidental tracking loss. We demonstrate the performance of our method in comparison to the state of the art in different challenging experiments including a popular public data set.

We present an approach for real-time, robust, and accurate hand pose estimation from moving egocentric RGB-D cameras in cluttered real environments. Existing methods typically fail for hand-object interactions in cluttered scenes imaged from egocentric viewpoints, common for virtual or augmented reality applications. Our approach uses two subsequently applied Convolutional Neural Networks (CNNs) to localize the hand and regress 3D joint locations. Hand localization is achieved by using a CNN to estimate the 2D position of the hand center in the input, even in the presence of clutter and occlusions. The localized hand position, together with the corresponding input depth value, is used to generate a normalized cropped image that is fed into a second CNN to regress relative 3D hand joint locations in real-time. For added accuracy, robustness, and temporal stability, we refine the pose estimates using a kinematic pose tracking energy. To train the CNNs, we introduce a new photorealistic dataset that uses a merged reality approach to capture and synthesize large amounts of annotated data of natural hand interaction in cluttered scenes. Through quantitative and qualitative evaluation, we show that our method is robust to self-occlusion and occlusions by objects, specifically in moving egocentric perspectives.

3D model-based gaze estimation methods are widely explored because of their good accuracy and ability to handle free head movement. Traditional methods with complex hardware systems (Eg. infrared lights, 3D sensors, etc.) are restricted to controlled environments, which significantly limit their practical utilities. In this paper, we propose a 3D model-based gaze estimation method with a single web-camera, which enables instant and portable eye gaze tracking. The key idea is to leverage on the proposed 3D eye-face model, from which we can estimate 3D eye gaze from observed 2D facial landmarks. The proposed system includes a 3D deformable eye-face model that is learned offline from multiple training subjects. Given the deformable model, individual 3D eye-face models and personal eye parameters can be recovered through the unified calibration algorithm. Experimental results show that the proposed method outperforms state-of-the-art methods while allowing convenient system setup and free head movement. A real time eye tracking system running at 30 FPS also validates the effectiveness and efficiency of the proposed method.

Traditional Structure-from-Motion (SfM) uses images captured by cameras as inputs. In this paper, we explore using light fields captured by plenoptic cameras or camera arrays as inputs. We call this solution plenoptic SfM or P-SfM solution. We first present a comprehensive theory on ray geometry transforms under light field pose variations. We derive the transforms of three typical ray manifolds: rays passing through a point or point-ray manifold, rays passing through a 3D line or ray-line manifold, and rays lying on a common 3D plane or ray-plane manifold. We show that by matching these manifolds across LFs, we can recover light field poses and conduct bundle adjustment in ray space. We validate our theory and framework on synthetic and real data on light fields of different scales: small scale LFs acquired using a LF camera and large scale LFs by a camera array. We show that our P-SfM technique can significantly improve the accuracy and reliability over regular SfM and PnP especially on traditionally challenging scenes where reliable feature point correspondences are difficult to obtain but line or plane correspondences are readily accessible.

This paper addresses the problem of converting a rasterized floorplan image into a vector-graphics representation. Unlike existing approaches that rely on a sequence of low-level image processing heuristics, we adopt a learning-based approach. A neural architecture first transforms a rasterized image to a set of junctions that represent low-level geometric and semantic information (e.g., wall corners or door end-points). Integer programming is then formulated to aggregate junctions into a set of simple primitives (e.g., wall lines, door lines, or icon boxes) to produce a vectorized floorplan, while ensuring a topologically and geometrically consistent result. Our algorithm significantly outperforms existing methods and achieves around 90% precision and recall, getting to the range of production-ready performance. The vector representation allows 3D model popup for better indoor scene visualization, direct model manipulation for architectural remodeling, and further computational applications such as data analysis. Our system is efficient: we have converted hundred thousand production-level floorplan images into the vector representation and generated 3D popup models.

We propose a no-reference image quality assessment (NR-IQA) approach that learns from rankings (RankIQA). To address the problem of limited IQA dataset size, we train a Siamese Network to rank images in terms of image quality by using synthetically generated distortions for which relative image quality is known. These ranked image sets can be automatically generated without laborious human labeling. We then use fine-tuning to transfer the knowledge represented in the trained Siamese Network to a traditional CNN that estimates absolute image quality from single images. We demonstrate how our approach can be made significantly more efficient than traditional Siamese Networks by forward propagating a batch of images through a single network and backpropagating gradients derived from all pairs of images in the batch. Experiments on the TID2013 benchmark show that we improve the state-of-the-art by over 5%. Furthermore, on the LIVE benchmark we show that our approach is superior to existing NR-IQA techniques and that we even outperform the state-of-the-art in full-reference IQA (FR-IQA) methods without having to resort to high-quality reference images to infer IQA.

Deep convolutional neural networks have achieved significant improvements on face recognition task due to their ability to learn highly discriminative features from tremendous amounts of face images. Many large scale face datasets exhibit long-tail distribution where a small number of entities (persons) have large number of face images while a large number of persons only have very few face samples (long tail). Most of the existing works alleviate this problem by simply cutting the tailed data and only keep identities with enough number of examples. Unlike these work, this paper investigated how long-tailed data impact the training of face CNNs and develop a novel loss function, called range loss, to effectively utilize the tailed data in training process. More specifically, range loss is designed to reduce overall intrapersonal variations while enlarge inter-personal differences simultaneously. Extensive experiments on two face recognition benchmarks, Labeled Faces in the Wild (LFW) and YouTube Faces (YTF), demonstrate the effectiveness of the proposed range loss in overcoming the long tail effect, and show the good generalization ability of the proposed methods.

Recent studies demonstrate the effectiveness of Recurrent Neural Networks (RNNs) for action recognition in videos. However, previous works mainly utilize video-level category as supervision to train RNNs, which may prohibit RNNs to learn complex motion structures along time. In this paper, we propose a recurrent pose-attention network (RPAN) to address this challenge, where we introduce a novel pose-attention mechanism to adaptively learn pose-related features at every time-step action prediction of RNNs. More specifically, we make three main contributions in this paper. Firstly, unlike previous works on pose-related action recognition, our RPAN is an end-to-end recurrent network which can exploit important spatial-temporal evolutions of human pose to assist action recognition in a unified framework. Secondly, instead of learning individual human-joint features separately, our pose-attention mechanism learns robust human-part features by sharing attention parameters partially on the semantically-related human joints. These human-part features are then fed into the human-part pooling layer to construct a highly-discriminative pose-related representation for temporal action modeling. Thirdly, one important byproduct of our RPAN is pose estimation in videos, which can be used for coarse pose annotation in action videos. We evaluate the proposed RPAN quantitatively and qualitatively on two popular benchmarks, i.e., Sub-JHMDB and PennAction. Experimental results show that RPAN outperforms the recent state-of-the-art methods on these challenging datasets.

Multi-person pose estimation in the wild is challenging. Although state-of-the-art human detectors have demonstrated good performance, small errors in localization and recognition are inevitable. These errors can cause failures for a single-person pose estimator (SPPE), especially for methods that solely depend on human detection results. In this paper, we propose a novel regional multi-person pose estimation (RMPE) framework to facilitate pose estimation in the presence of inaccurate human bounding boxes. Our framework consists of three components: Symmetric Spatial Transformer Network (SSTN), Parametric Pose Non-Maximum-Suppression (NMS), and Pose-Guided Proposals Generator (PGPG). Our method is able to handle inaccurate bounding boxes and redundant detections, allowing it to achieve 76.7 mAP on the MPII (multi person) dataset. Our model and source codes are made publicly available.