【计算机视觉论文速递】2018-12-24~12-28
- 2018-12-24
- 2018-12-25
- 2018-12-26
- 2018-12-27
- 2018-12-28
本文分享共计39篇论文，涉及图像分类、目标检测、语义分割、GAN、姿态估计、SLAM、显著性目标检测和Zero-Shot Learning等方向。
[TOC]
《Slimmable Neural Networks》
ICLR 2019
arXiv：https://arxiv.org/abs/1812.08928
github：https://github.com/JiahuiYu/slimmable_networks
We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and in many cases better) ImageNet classification accuracy than individually trained models of MobileNet v1, MobileNet v2, ShuffleNet and ResNet-50 at different widths respectively. We also demonstrate better performance of slimmable models compared with individual ones across a wide range of applications including COCO bounding-box object detection, instance segmentation and person keypoint detection without tuning hyper-parameters. Lastly we visualize and discuss the learned features of slimmable networks. Code and models are available at: https://github.com/JiahuiYu/slimmable_networks
《ChamNet: Towards Efficient Network Design through Platform-Aware Model Adaptation》
arXiv：https://arxiv.org/abs/1812.08934
This paper proposes an efficient neural network (NN) architecture design methodology called Chameleon that honors given resource constraints. Instead of developing new building blocks or using computationally-intensive reinforcement learning algorithms, our approach leverages existing efficient network building blocks and focuses on exploiting hardware traits and adapting computation resources to fit target latency and/or energy constraints. We formulate platform-aware NN architecture search in an optimization framework and propose a novel algorithm to search for optimal architectures aided by efficient accuracy and resource (latency and/or energy) predictors. At the core of our algorithm lies an accuracy predictor built atop Gaussian Process with Bayesian optimization for iterative sampling. With a one-time building cost for the predictors, our algorithm produces state-of-the-art model architectures on different platforms under given constraints in just minutes. Our results show that adapting computation resources to building blocks is critical to model performance. Without the addition of any bells and whistles, our models achieve significant accuracy improvements against state-of-the-art hand-crafted and automatically designed architectures. We achieve 73.8% and 75.3% top-1 accuracy on ImageNet at 20ms latency on a mobile CPU and DSP. At reduced latency, our models achieve up to 8.5% (4.8%) and 6.6% (9.3%) absolute top-1 accuracy improvements compared to MobileNetV2 and MnasNet, respectively, on a mobile CPU (DSP), and 2.7% (4.6%) and 5.6% (2.6%) accuracy gains over ResNet-101 and ResNet-152, respectively, on an Nvidia GPU (Intel CPU).
《Learning from Web Data: the Benefit of Unsupervised Object Localization》
IEEE TIP 2019
arXiv：https://arxiv.org/abs/1812.09232
Annotating a large number of training images is very time-consuming. In this background, this paper focuses on learning from easy-to-acquire web data and utilizes the learned model for fine-grained image classification in labeled datasets. Currently, the performance gain from training with web data is incremental, like a common saying "better than nothing, but not by much". Conventionally, the community looks to correcting the noisy web labels to select informative samples. In this work, we first systematically study the built-in gap between the web and standard datasets, i.e. different data distributions between the two kinds of data. Then, in addition to using web labels, we present an unsupervised object localization method, which provides critical insights into the object density and scale in web images. Specifically, we design two constraints on web data to substantially reduce the difference of data distributions for the web and standard datasets. First, we present a method to control the scale, localization and number of objects in the detected region. Second, we propose to select the regions containing objects that are consistent with the web tag. Based on the two constraints, we are able to process web images to reduce the gap, and the processed web data is used to better assist the standard dataset to train CNNs. Experiments on several fine-grained image classification datasets confirm that our method performs favorably against the state-of-the-art methods.
《Chinese Herbal Recognition based on Competitive Attentional Fusion of Multi-hierarchies Pyramid Features》
arXiv：https://arxiv.org/abs/1812.09648
github：https://github.com/scut-aitcm/Chinese-Herbs-Dataset
注：中药识别，有点意思的
Convolution neural netwotks (CNNs) are successfully applied in image recognition task. In this study, we explore the approach of automatic herbal recognition with CNNs and build the standard Chinese herbs datasets firstly. According to the characteristics of herbal images, we proposed the competitive attentional fusion pyramid networks to model the features of herbal image, which mdoels the relationship of feature maps from different levels, and re-weights multi-level channels with channel-wise attention mechanism. In this way, we can dynamically adjust the weight of feature maps from various layers, according to the visual characteristics of each herbal image. Moreover, we also introduce the spatial attention to recalibrate the misaligned features caused by sampling in features amalgamation. Extensive experiments are conducted on our proposed datasets and validate the superior performance of our proposed models. The Chinese herbs datasets will be released upon acceptance to facilitate the research of Chinese herbal recognition.
《A Survey to Deep Facial Attribute Analysis》
arXiv：https://arxiv.org/abs/1812.10265
注：好综述
Facial attribute analysis has received considerable attention with the development of deep neural networks in the past few years. Facial attribute analysis contains two crucial issues: Facial Attribute Estimation (FAE), which recognizes whether facial attributes are present in given images, and Facial Attribute Manipulation (FAM), which synthesizes or removes desired facial attributes. In this paper, we provide a comprehensive survey on deep facial attribute analysis covering FAE and FAM. First, we present the basic knowledge of the two stages (i.e., data pre-processing and model construction) in the general deep facial attribute analysis pipeline. Second, we summarize the commonly used datasets and performance metrics. Third, we create a taxonomy of the state-of-the-arts and review detailed algorithms in FAE and FAM, respectively. Furthermore, we introduce several additional facial attribute related issues and applications. Finally, the possible challenges and future research directions are discussed.
《A Smart Security System with Face Recognition》
arXiv：https://arxiv.org/abs/1812.09127
Web-based technology has improved drastically in the past decade. As a result, security technology has become a major help to protect our daily life. In this paper, we propose a robust security based on face recognition system (SoF). In particular, we develop this system to giving access into a home for authenticated users. The classifier is trained by using a new adaptive learning method. The training data are initially collected from social networks. The accuracy of the classifier is incrementally improved as the user starts using the system. A novel method has been introduced to improve the classifier model by human interaction and social media. By using a deep learning framework - TensorFlow, it will be easy to reuse the framework to adopt with many devices and applications.
《3D multirater RCNN for multimodal multiclass detection and characterisation of extremely small objects》
MIDL 2019 submission
arXiv：https://arxiv.org/abs/1812.09046
Extremely small objects (ESO) have become observable on clinical routine magnetic resonance imaging acquisitions, thanks to a reduction in acquisition time at higher resolution. Despite their small size (usually <10 voxels per object for an image of more than 106 voxels), these markers reflect tissue damage and need to be accounted for to investigate the complete phenotype of complex pathological pathways. In addition to their very small size, variability in shape and appearance leads to high labelling variability across human raters, resulting in a very noisy gold standard. Such objects are notably present in the context of cerebral small vessel disease where enlarged perivascular spaces and lacunes, commonly observed in the ageing population, are thought to be associated with acceleration of cognitive decline and risk of dementia onset. In this work, we redesign the RCNN model to scale to 3D data, and to jointly detect and characterise these important markers of age-related neurovascular changes. We also propose training strategies enforcing the detection of extremely small objects, ensuring a tractable and stable training process.
《Detection of distal radius fractures trained by a small set of X-ray images and Faster R-CNN》
arXiv：https://arxiv.org/abs/1812.09025
Distal radius fractures are the most common fractures of the upper extremity in humans. As such, they account for a significant portion of the injuries that present to emergency rooms and clinics throughout the world. We trained a Faster R-CNN, a machine vision neural network for object detection, to identify and locate distal radius fractures in anteroposterior X-ray images. We achieved an accuracy of 96% in identifying fractures and mean Average Precision, mAP, of 0.866. This is significantly more accurate than the detection achieved by physicians and radiologists. These results were obtained by training the deep learning network with only 38 original images of anteroposterior hands X-ray images with fractures. This opens the possibility to detect with this type of neural network rare diseases or rare symptoms of common diseases , where only a small set of diagnosed X-ray images could be collected for each disease.
《Practical Adversarial Attack Against Object Detector》
arXiv：https://arxiv.org/abs/1812.10217
注：很有意思的研究
In this paper, we proposed the first practical adversarial attacks against object detectors in realistic situations: the adversarial examples are placed in different angles and distances, especially in the long distance (over 20m) and wide angles 120 degree. To improve the robustness of adversarial examples, we proposed the nested adversarial examples and introduced the image transformation techniques. Transformation methods aim to simulate the variance factors such as distances, angles, illuminations, etc., in the physical world. Two kinds of attacks were implemented on YOLO V3, a state-of-the-art real-time object detector: hiding attack that fools the detector unable to recognize the object, and appearing attack that fools the detector to recognize the non-existent object. The adversarial examples are evaluated in three environments: indoor lab, outdoor environment, and the real road, and demonstrated to achieve the success rate up to 92.4% based on the distance range from 1m to 25m. In particular, the real road testing of hiding attack on a straight road and a crossing road produced the success rate of 75% and 64% respectively, and the appearing attack obtained the success rates of 63% and 81% respectively, which we believe, should catch the attention of the autonomous driving community.
《Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes》
arXiv：https://arxiv.org/abs/1812.09953
During the last half decade, convolutional neural networks (CNNs) have triumphed over semantic segmentation, which is one of the core tasks in many applications such as autonomous driving and augmented reality. However, to train CNNs requires a considerable amount of data, which is difficult to collect and laborious to annotate. Recent advances in computer graphics make it possible to train CNNs on photo-realistic synthetic imagery with computer-generated annotations. Despite this, the domain mismatch between the real images and the synthetic data hinders the models' performance. Hence, we propose a curriculum-style learning approach to minimizing the domain gap in urban scene semantic segmentation. The curriculum domain adaptation solves easy tasks first to infer necessary properties about the target domain; in particular, the first task is to learn global label distributions over images and local distributions over landmark superpixels. These are easy to estimate because images of urban scenes have strong idiosyncrasies (e.g., the size and spatial relations of buildings, streets, cars, etc.). We then train a segmentation network, while regularizing its predictions in the target domain to follow those inferred properties. In experiments, our method outperforms the baselines on two datasets and two backbone networks. We also report extensive ablation studies about our approach.
《Improving MMD-GAN Training with Repulsive Loss Function》
ICLR 2019
arXiv：https://arxiv.org/abs/1812.09916
Generative adversarial nets (GANs) are widely used to learn the data sampling process and their performance may heavily depend on the loss functions, given a limited computational budget. This study revisits MMD-GAN that uses the maximum mean discrepancy (MMD) as the loss function for GAN and makes two contributions. First, we argue that the existing MMD loss function may discourage the learning of fine details in data as it attempts to contract the discriminator outputs of real data. To address this issue, we propose a repulsive loss function to actively learn the difference among the real data by simply rearranging the terms in MMD. Second, inspired by the hinge loss, we propose a bounded Gaussian kernel to stabilize the training of MMD-GAN with the repulsive loss function. The proposed methods are applied to the unsupervised image generation tasks on CIFAR-10, STL-10, CelebA, and LSUN bedroom datasets. Results show that the repulsive loss function significantly improves over the MMD loss at no additional computational cost and outperforms other representative loss functions. The proposed methods achieve an FID score of 16.21 on the CIFAR-10 dataset using a single DCGAN network and spectral normalization.
《Saliency Guided Hierarchical Robust Visual Tracking》
arXiv：https://arxiv.org/abs/1812.08973
A saliency guided hierarchical visual tracking (SHT) algorithm containing global and local search phases is proposed in this paper. In global search, a top-down saliency model is novelly developed to handle abrupt motion and appearance variation problems. Nineteen feature maps are extracted first and combined with online learnt weights to produce the final saliency map and estimated target locations. After the evaluation of integration mechanism, the optimum candidate patch is passed to the local search. In local search, a superpixel based HSV histogram matching is performed jointly with an L2-RLS tracker to take both color distribution and holistic appearance feature of the object into consideration. Furthermore, a linear refinement search process with fast iterative solver is implemented to attenuate the possible negative influence of dominant particles. Both qualitative and quantitative experiments are conducted on a series of challenging image sequences. The superior performance of the proposed method over other state-of-the-art algorithms is demonstrated by comparative study.
《Perceptually-based single-image depth super-resolution》
arXiv：https://arxiv.org/abs/1812.09874
RGBD images, combining high-resolution color and lower-resolution depth from various types of depth sensors, are increasingly common. One can significantly improve the resolution of depth images by taking advantage of color information; deep learning methods make combining color and depth information particularly easy. However, fusing these two sources of data may lead to a variety of artifacts. If depth maps are used to reconstruct 3D shapes, e.g., for virtual reality applications, the visual quality of upsampled images is particularly important. To achieve high-quality results, visual metric need to be taken into account. The main idea of our approach is to measure the quality of depth map upsampling using renderings of resulting 3D surfaces. We demonstrate that a simple visual appearance-based loss, when used with either a trained CNN or simply a deep prior, yields significantly improved 3D shapes, as measured by a number of existing perceptual metrics. We compare this approach with a number of existing optimization and learning-based techniques.
《A Survey on Non-rigid 3D Shape Analysis》
arXiv：https://arxiv.org/abs/1812.10111
Shape is an important physical property of natural and manmade 3D objects that characterizes their external appearances. Understanding differences between shapes and modeling the variability within and across shape classes, hereinafter referred to as \emph{shape analysis}, are fundamental problems to many applications, ranging from computer vision and computer graphics to biology and medicine. This chapter provides an overview of some of the recent techniques that studied the shape of 3D objects that undergo non-rigid deformations including bending and stretching. Recent surveys that covered some aspects such classification, retrieval, recognition, and rigid or nonrigid registration, focused on methods that use shape descriptors. Descriptors, however, provide abstract representations that do not enable the exploration of shape variability. In this chapter, we focus on recent techniques that treated the shape of 3D objects as points in some high dimensional space where paths describe deformations. Equipping the space with a suitable metric enables the quantification of the range of deformations of a given shape, which in turn enables (1) comparing and classifying 3D objects based on their shape, (2) computing smooth deformations, i.e. geodesics, between pairs of objects, and (3) modeling and exploring continuous shape variability in a collection of 3D models. This article surveys and classifies recent developments in this field, outlines fundamental issues, discusses their potential applications in computer vision and graphics, and highlights opportunities for future research. Our primary goal is to bridge the gap between various techniques that have been often independently proposed by different communities including mathematics and statistics, computer vision and graphics, and medical image analysis.
《Learning a Disentangled Embedding for Monocular 3D Shape Retrieval and Pose Estimation》
arXiv：https://arxiv.org/abs/1812.09899
We propose a novel approach to jointly perform 3D object retrieval and pose estimation from monocular images.In order to make the method robust to real world scene variations in the images, e.g. texture, lighting and background,we learn an embedding space from 3D data that only includes the relevant information, namely the shape and pose.Our method can then be trained for robustness under real world scene variations without having to render a large training set simulating these variations. Our learned embedding explicitly disentangles a shape vector and a pose vector, which alleviates both pose bias for 3D shape retrieval and categorical bias for pose estimation. Having the learned disentangled embedding, we train a CNN to map the images to the embedding space, and then retrieve the closest 3D shape from the database and estimate the 6D pose of the object using the embedding vectors. Our method achieves 10.8 median error for pose estimation and 0.514 top-1-accuracy for category agnostic 3D object retrieval on the Pascal3D+ dataset. It therefore outperforms the previous state-of-the-art methods on both tasks.
《Structure-Aware 3D Hourglass Network for Hand Pose Estimation from Single Depth Image》
BMVC 2018
arXiv：https://arxiv.org/abs/1812.10320
In this paper, we propose a novel structure-aware 3D hourglass network for hand pose estimation from a single depth image, which achieves state-of-the-art results on MSRA and NYU datasets. Compared to existing works that perform image-to-coordination regression, our network takes 3D voxel as input and directly regresses 3D heatmap for each joint. To be specific, we use hourglass network as our backbone network and modify it into 3D form. We explicitly model tree-like finger bone into the network as well as in the loss function in an end-to-end manner, in order to take the skeleton constraints into consideration. Final estimation can then be easily obtained from voxel density map with simple post-processing. Experimental results show that the proposed structure-aware 3D hourglass network is able to achieve a mean joint error of 7.4 mm in MSRA and 8.9 mm in NYU datasets, respectively.
《TextNet: Irregular Text Reading from Images with an End-to-End Trainable Network》
ACCV 2018 oral
百度出品，必属精品，Mark！
arXiv：https://arxiv.org/abs/1812.09900
Reading text from images remains challenging due to multi-orientation, perspective distortion and especially the curved nature of irregular text. Most of existing approaches attempt to solve the problem in two or multiple stages, which is considered to be the bottleneck to optimize the overall performance. To address this issue, we propose an end-to-end trainable network architecture, named TextNet, which is able to simultaneously localize and recognize irregular text from images. Specifically, we develop a scale-aware attention mechanism to learn multi-scale image features as a backbone network, sharing fully convolutional features and computation for localization and recognition. In text detection branch, we directly generate text proposals in quadrangles, covering oriented, perspective and curved text regions. To preserve text features for recognition, we introduce a perspective RoI transform layer, which can align quadrangle proposals into small feature maps. Furthermore, in order to extract effective features for recognition, we propose to encode the aligned RoI features by RNN into context information, combining spatial attention mechanism to generate text sequences. This overall pipeline is capable of handling both regular and irregular cases. Finally, text localization and recognition tasks can be jointly trained in an end-to-end fashion with designed multi-task loss. Experiments on standard benchmarks show that the proposed TextNet can achieve state-of-the-art performance, and outperform existing approaches on irregular datasets by a large margin.
《SMILER: Saliency Model Implementation Library for Experimental Research》
arXiv：https://arxiv.org/abs/1812.08848
github：https://github.com/TsotsosLab/SMILER
The Saliency Model Implementation Library for Experimental Research (SMILER) is a new software package which provides an open, standardized, and extensible framework for maintaining and executing computational saliency models. This work drastically reduces the human effort required to apply saliency algorithms to new tasks and datasets, while also ensuring consistency and procedural correctness for results and conclusions produced by different parties. At its launch SMILER already includes twenty three saliency models (fourteen models based in MATLAB and nine supported through containerization), and the open design of SMILER encourages this number to grow with future contributions from the community. The project may be downloaded and contributed to through its GitHub page: https://github.com/TsotsosLab/SMILER
《A Unified Framework for Mutual Improvement of SLAM and Semantic Segmentation》
arXiv：https://arxiv.org/abs/1812.10016
This paper presents a novel framework for simultaneously implementing localization and segmentation, which are two of the most important vision-based tasks for robotics. While the goals and techniques used for them were considered to be different previously, we show that by making use of the intermediate results of the two modules, their performance can be enhanced at the same time. Our framework is able to handle both the instantaneous motion and long-term changes of instances in localization with the help of the segmentation result, which also benefits from the refined 3D pose information. We conduct experiments on various datasets, and prove that our framework works effectively on improving the precision and robustness of the two tasks and outperforms existing localization and segmentation algorithms.
《Selectivity or Invariance: Boundary-aware Salient Object Detection》
arXiv：https://arxiv.org/abs/1812.10066
Typically, a salient object detection (SOD) model faces opposite requirements in processing object interiors and boundaries. The features of interiors should be invariant to strong appearance change so as to pop-out the salient object as a whole, while the features of boundaries should be selective to slight appearance change to distinguish salient objects and background. To address this selectivity-invariance dilemma, we propose a novel boundary-aware network with successive dilation for image-based SOD. In this network, the feature selectivity at boundaries is enhanced by incorporating a boundary localization stream, while the feature invariance at interiors is guaranteed with a complex interior perception stream. Moreover, a transition compensation stream is adopted to amend the probable failures in transitional regions between interiors and boundaries. In particular, an integrated successive dilation module is proposed to enhance the feature invariance at interiors and transitional regions. Extensive experiments on four datasets show that the proposed approach outperforms 11 state-of-the-art methods.
《3D PersonVLAD: Learning Deep Global Representations for Video-based Person Re-identification》
IEEE Transactions on Neural Networks and Learning Systems
arXiv：https://arxiv.org/abs/1812.10222
In this paper, we introduce a global video representation to video-based person re-identification (re-ID) that aggregates local 3D features across the entire video extent. Most of the existing methods rely on 2D convolutional networks (ConvNets) to extract frame-wise deep features which are pooled temporally to generate the video-level representations. However, 2D ConvNets lose temporal input information immediately after the convolution, and a separate temporal pooling is limited in capturing human motion in shorter sequences. To this end, we present a \textit{global} video representation (3D PersonVLAD), complementary to 3D ConvNets as a novel layer to capture the appearance and motion dynamics in full-length videos. However, encoding each video frame in its entirety and computing an aggregate global representation across all frames is tremendously challenging due to occlusions and misalignments. To resolve this, our proposed network is further augmented with 3D part alignment module to learn local features through soft-attention module. These attended features are statistically aggregated to yield identity-discriminative representations. Our global 3D features are demonstrated to achieve state-of-the-art results on three benchmark datasets: MARS \cite{MARS}, iLIDS-VID \cite{VideoRanking}, and PRID 2011
《Spatial and Temporal Mutual Promotion for Video-based Person Re-identification》
AAAI 2019
arXiv：https://arxiv.org/abs/1812.10305
Video-based person re-identification is a crucial task of matching video sequences of a person across multiple camera views. Generally, features directly extracted from a single frame suffer from occlusion, blur, illumination and posture changes. This leads to false activation or missing activation in some regions, which corrupts the appearance and motion representation. How to explore the abundant spatial-temporal information in video sequences is the key to solve this problem. To this end, we propose a Refining Recurrent Unit (RRU) that recovers the missing parts and suppresses noisy parts of the current frame's features by referring historical frames. With RRU, the quality of each frame's appearance representation is improved. Then we use the Spatial-Temporal clues Integration Module (STIM) to mine the spatial-temporal information from those upgraded features. Meanwhile, the multi-level training objective is used to enhance the capability of RRU and STIM. Through the cooperation of those modules, the spatial and temporal features mutually promote each other and the final spatial-temporal feature representation is more discriminative and robust. Extensive experiments are conducted on three challenging datasets, i.e., iLIDS-VID, PRID-2011 and MARS. The experimental results demonstrate that our approach outperforms existing state-of-the-art methods of video-based person re-identification on iLIDS-VID and MARS and achieves favorable results on PRID-2011.
《Cluster Loss for Person Re-Identification》
ICVGIP 2018
arXiv：https://arxiv.org/abs/1812.10325
Person re-identification (ReID) is an important problem in computer vision, especially for video surveillance applications. The problem focuses on identifying people across different cameras or across different frames of the same camera. The main challenge lies in identifying the similarity of the same person against large appearance and structure variations, while differentiating between individuals. Recently, deep learning networks with triplet loss have become a common framework for person ReID. However, triplet loss focuses on obtaining correct orders on the training set. We demonstrate that it performs inferior in a clustering task. In this paper, we design a cluster loss, which can lead to the model output with a larger inter-class variation and a smaller intra-class variation compared to the triplet loss. As a result, our model has a better generalization ability and can achieve higher accuracy on the test set especially for a clustering task. We also introduce a batch hard training mechanism for improving the results and faster convergence of training.
《EgoReID: Person re-identification in Egocentric Videos Acquired by Mobile Devices with First-Person Point-of-View》
arXiv：https://arxiv.org/abs/1812.09570
Widespread use of wearable cameras and recording devices such as cellphones have opened the door to a lot of interesting research in first-person Point-of-view (POV) videos (egocentric videos). In recent years, we have seen the performance of video-based person Re-Identification (ReID) methods improve considerably. However, with the influx of varying video domains, such as egocentric videos, it has become apparent that there are still many open challenges to be faced. These challenges are a result of factors such as poor video quality due to ego-motion, blurriness, severe changes in lighting conditions and perspective distortions. To facilitate the research towards conquering these challenges, this paper contributes a new, first-of-its-kind dataset called EgoReID. The dataset is captured using 3 mobile cellphones with non-overlapping field-of-view. It contains 900 IDs and around 10,200 tracks with a total of 176,000 detections. Moreover, for each video we also provide 12-sensor meta data. Directly applying current approaches to our dataset results in poor performance. Considering the unique nature of our dataset, we propose a new framework which takes advantage of both visual and sensor meta data to successfully perform Person ReID. In this paper, we propose to adopt human body region parsing to extract local features from different body regions and then employ 3D convolution to better encode temporal information of each sequence of body parts. In addition, we also employ sensor meta data to determine target's next camera and their estimated time of arrival, such that the search is only performed among tracks present in the predicted next camera around the estimated time. This considerably improves our ReID performance as it significantly reduces our search space.
《3DSRnet: Video Super-resolution using 3D Convolutional Neural Networks》
arXiv：https://arxiv.org/abs/1812.09079
In video super-resolution, the spatio-temporal coherence between, and among the frames must be exploited appropriately for accurate prediction of the high resolution frames. Although 2D convolutional neural networks (CNNs) are powerful in modelling images, 3D-CNNs are more suitable for spatio-temporal feature extraction as they can preserve temporal information. To this end, we propose an effective 3D-CNN for video super-resolution, called the 3DSRnet that does not require motion alignment as preprocessing. Our 3DSRnet maintains the temporal depth of spatio-temporal feature maps to maximally capture the temporally nonlinear characteristics between low and high resolution frames, and adopts residual learning in conjunction with the sub-pixel outputs. It outperforms the most state-of-the-art method with average 0.45 and 0.36 dB higher in PSNR for scales 3 and 4, respectively, in the Vidset4 benchmark. Our 3DSRnet first deals with the performance drop due to scene change, which is important in practice but has not been previously considered.
《A Multiscale Image Denoising Algorithm Based On Dilated Residual Convolution Network》
arXiv：https://arxiv.org/abs/1812.09131
Image denoising is a classical problem in low level computer vision. Model-based optimization methods and deep learning approaches have been the two main strategies for solving the problem. Model-based optimization methods are flexible for handling different inverse problems but are usually time-consuming. In contrast, deep learning methods have fast testing speed but the performance of these CNNs is still inferior. To address this issue, here we propose a novel deep residual learning model that combines the dilated residual convolution and multi-scale convolution groups. Due to the complex patterns and structures of inside an image, the multiscale convolution group is utilized to learn those patterns and enlarge the receptive field. Specifically, the residual connection and batch normalization are utilized to speed up the training process and maintain the denoising performance. In order to decrease the gridding artifacts, we integrate the hybrid dilated convolution design into our model. To this end, this paper aims to train a lightweight and effective denoiser based on multiscale convolution group. Experimental results have demonstrated that the enhanced denoiser can not only achieve promising denoising results, but also become a strong competitor in practical application.
《Domain-Aware Generalized Zero-Shot Learning》
arXiv：https://arxiv.org/abs/1812.09903
Generalized zero-shot learning (GZSL) is the problem of learning a classifier where some classes have samples, and others are learned from side information, like semantic attributes or text description, in a zero-shot learning fashion (ZSL). A major challenge in GZSL is to learn consistently for those two different domains. Here we describe a probabilistic approach that breaks the model into three modular components, and then combines them in a consistent way. Specifically, our model consists of three classifiers: A "gating" model that softly decides if a sample is from a "seen" class and two experts: a ZSL expert, and an expert model for seen classes. We address two main difficulties in this approach: How to provide an accurate estimate of the gating probability without any training samples for unseen classes; and how to use an expert predictions when it observes samples outside of its domain. The key insight in our approach is to pass information between the three models to improve each others accuracy, while keeping the modular structure. We test our approach, Domain-Aware GZSL (DAZL) on three standard GZSL benchmark datasets (AWA, CUB, SUN), and find that it largely outperforms state-of-the-art GZSL models. DAZL is also the first model that closes the gap and surpasses the performance of generative models for GZSL, even-though it is a light-weight model that is much easier to train and tune.
《Learning Compositional Representations for Few-Shot Recognition》
arXiv：https://arxiv.org/abs/1812.09213
One of the key limitations of modern deep learning based approaches lies in the amount of data required to train them. Humans, on the other hand, can learn to recognize novel categories from just a few examples. Instrumental to this rapid learning ability is the compositional structure of concept representations in the human brain - something that deep learning models are lacking. In this work we make a step towards bridging this gap between human and machine learning by introducing a simple regularization technique that allows the learned representation to be decomposable into parts. We evaluate the proposed approach on three datasets: CUB-200-2011, SUN397, and ImageNet, and demonstrate that our compositional representations require fewer examples to learn classifiers for novel categories, outperforming state-of-the-art few-shot learning approaches by a significant margin.
《Similarity R-C3D for Few-shot Temporal Activity Detection》
arXiv：https://arxiv.org/abs/1812.10000
Many activities of interest are rare events, with only a few labeled examples available. Therefore models for temporal activity detection which are able to learn from a few examples are desirable. In this paper, we present a conceptually simple and general yet novel framework for few-shot temporal activity detection which detects the start and end time of the few-shot input activities in an untrimmed video. Our model is end-to-end trainable and can benefit from more few-shot examples. At test time, each proposal is assigned the label of the few-shot activity class corresponding to the maximum similarity score. Our Similarity R-C3D method outperforms previous work on three large-scale benchmarks for temporal activity detection (THUMOS14, ActivityNet1.2, and ActivityNet1.3 datasets) in the few-shot setting. Our code will be made available.
《A Scale Invariant Approach for Sparse Signal Recovery》
arXiv：https://arxiv.org/abs/1812.08852
In this paper, we study the ratio of the L1 and L2 norms, denoted as L1/L2, to promote sparsity. Due to the non-convexity and non-linearity, there has been little attention to this scale-invariant metric. Compared to popular models in the literature such as the Lp model for p∈(0,1) and the transformed L1 (TL1), this ratio model is parameter free. Theoretically, we present a weak null space property (wNSP) and prove that any sparse vector is a local minimizer of the L1/L2 model provided with this wNSP condition. Computationally, we focus on a constrained formulation that can be solved via the alternating direction method of multipliers (ADMM). Experiments show that the proposed approach is comparable to the state-of-the-art methods in sparse recovery. In addition, a variant of the L1/L2 model to apply on the gradient is also discussed with a proof-of-concept example of MRI reconstruction.construction.
《Polygonal approximation of digital planar curve using novel significant measure》
arXiv：https://arxiv.org/abs/1812.09271
This paper presents an iterative smoothing technique for polygonal approximation of digital image boundary. The technique starts with finest initial segmentation points of a curve. The contribution of initially segmented points towards preserving the original shape of the image boundary is determined by computing the significant measure of every initial segmentation points which is sensitive to sharp turns, which may be missed easily when conventional significant measures are used for detecting dominant points. The proposed method differentiates between the situations when a point on the curve between two points on a curve projects directly upon the line segment or beyond this line segment. It not only identifies these situations, but also computes its significant contribution for these situations differently. This situation-specific treatment allows preservation of points with high curvature even as revised set of dominant points are derived. The experimental results show that the proposed technique competes well with the state of the art techniques.
《Quicker ADC : Unlocking the hidden potential of Product Quantization with SIMD》
arXiv：https://arxiv.org/abs/1812.09162
github：https://github.com/technicolor-research/faiss-quickeradc
Efficient Nearest Neighbor (NN) search in high-dimensional spaces is a foundation of many multimedia retrieval systems. A common approach is to rely on Product Quantization that allows storing large vector databases in memory and also allows efficient distance computations. Yet, implementations of nearest neighbor search with Product Quantization have their performance limited by the many memory accesses they perform. Following this observation, André et al. proposed more efficient implementations of m×4 product quantizers (PQ) leveraging specific SIMD instructions. Quicker ADC contributes additional implementations not limited to m×4 codes and relying on AVX-512, the latest revision of SIMD instruction set. In doing so, Quicker ADC faces the challenge of using efficiently 5,6 and 7-bit shuffles that do not align to computer bytes or words. To this end, we introduce (i) irregular product quantizers combining sub-quantizers of different granularity and (ii) split tables allowing lookup tables larger than registers. We evaluate Quicker ADC with multiple indexes including Inverted Multi-Indexes and IVF HNSW and show that it outperforms FAISS PQ implementation and optimization (i.e., Polysemous codes) for numerous configurations. Finally, we open-source at this http URL a fork of FAISS that includes Quicker ADC.
《Writer-Aware CNN for Parsimonious HMM-Based Offline Handwritten Chinese Text Recognition》
arXiv：https://arxiv.org/abs/1812.09809
Recently, the hybrid convolutional neural network hidden Markov model (CNN-HMM) has been introduced for offline handwritten Chinese text recognition (HCTR) and has achieved state-of-the-art performance. In a CNN-HMM system, a handwritten text line is modeled by a series of cascading HMMs, each representing one character, and the posterior distributions of HMM states are calculated by CNN. However, modeling each of the large vocabulary of Chinese characters with a uniform and fixed number of hidden states requires high memory and computational costs and makes the tens of thousands of HMM state classes confusing. Another key issue of CNN-HMM for HCTR is the diversified writing style, which leads to model strain and a significant performance decline for specific writers. To address these issues, we propose a writer-aware CNN based on parsimonious HMM (WCNN-PHMM). Validated on the ICDAR 2013 competition of CASIA-HWDB database, the more compact WCNN-PHMM of a 7360-class vocabulary can achieve a relative character error rate (CER) reduction of 16.6% over the conventional CNN-HMM without considering language modeling. Moreover, the state-tying results of PHMM explicitly show the information sharing among similar characters and the confusion reduction of tied state classes. Finally, we visualize the learned writer codes and demonstrate the strong relationship with the writing styles of different writers. To the best of our knowledge, WCNN-PHMM yields the best results on the ICDAR 2013 competition set, demonstrating its power when enlarging the size of the character vocabulary.
《Color Image Enhancement Method Based on Weighted Image Guided Filtering》
arXiv：https://arxiv.org/abs/1812.09930
A novel color image enhancement method is proposed based on Retinex to enhance color images under non-uniform illumination or poor visibility conditions. Different from the conventional Retinex algorithms, the Weighted Guided Image Filter is used as a surround function instead of the Gaussian filter to estimate the background illumination, which can overcome the drawbacks of local blur and halo artifact that may appear by Gaussian filter. To avoid color distortion, the image is converted to the HSI color model, and only the intensity channel is enhanced. Then a linear color restoration algorithm is adopted to convert the enhanced intensity image back to the RGB color model, which ensures the hue is constant and undistorted. Experimental results show that the proposed method is effective to enhance both color and gray images with low exposure and non-uniform illumination, resulting in better visual quality than traditional method. At the same time, the objective evaluation indicators are also superior to the conventional methods. In addition, the efficiency of the proposed method is also improved thanks to the linear color restoration algorithm.
《Coupled Recurrent Network (CRN)》
arXiv：https://arxiv.org/abs/1812.10071
Many semantic video analysis tasks can benefit from multiple, heterogenous signals. For example, in addition to the original RGB input sequences, sequences of optical flow are usually used to boost the performance of human action recognition in videos. To learn from these heterogenous input sources, existing methods reply on two-stream architectural designs that contain independent, parallel streams of Recurrent Neural Networks (RNNs). However, two-stream RNNs do not fully exploit the reciprocal information contained in the multiple signals, let alone exploit it in a recurrent manner. To this end, we propose in this paper a novel recurrent architecture, termed Coupled Recurrent Network (CRN), to deal with multiple input sources. In CRN, the parallel streams of RNNs are coupled together. Key design of CRN is a Recurrent Interpretation Block (RIB) that supports learning of reciprocal feature representations from multiple signals in a recurrent manner. Different from RNNs which stack the training loss at each time step or the last time step, we propose an effective and efficient training strategy for CRN. Experiments show the efficacy of the proposed CRN. In particular, we achieve the new state of the art on the benchmark datasets of human action recognition and multi-person pose estimation.
《RegNet: Learning the Optimization of Direct Image-to-Image Pose Registration》
arXiv：https://arxiv.org/abs/1812.10212
Direct image-to-image alignment that relies on the optimization of photometric error metrics suffers from limited convergence range and sensitivity to lighting conditions. Deep learning approaches has been applied to address this problem by learning better feature representations using convolutional neural networks, yet still require a good initialization. In this paper, we demonstrate that the inaccurate numerical Jacobian limits the convergence range which could be improved greatly using learned approaches. Based on this observation, we propose a novel end-to-end network, RegNet, to learn the optimization of image-to-image pose registration. By jointly learning feature representation for each pixel and partial derivatives that replace handcrafted ones (e.g., numerical differentiation) in the optimization step, the neural network facilitates end-to-end optimization. The energy landscape is constrained on both the feature representation and the learned Jacobian, hence providing more flexibility for the optimization as a consequence leads to more robust and faster convergence. In a series of experiments, including a broad ablation study, we demonstrate that RegNet is able to converge for large-baseline image pairs with fewer iterations.
《Learning Not to Learn: Training Deep Neural Networks with Biased Data》
arXiv：https://arxiv.org/abs/1812.10352
We propose a novel regularization algorithm to train deep neural networks, in which data at training time is severely biased. Since a neural network efficiently learns data distribution, a network is likely to learn the bias information to categorize input data. It leads to poor performance at test time, if the bias is, in fact, irrelevant to the categorization. In this paper, we formulate a regularization loss based on mutual information between feature embedding and bias. Based on the idea of minimizing this mutual information, we propose an iterative algorithm to unlearn the bias information. We employ an additional network to predict the bias distribution and train the network adversarially against the feature embedding network. At the end of learning, the bias prediction network is not able to predict the bias not because it is poorly trained, but because the feature embedding network successfully unlearns the bias information. We also demonstrate quantitative and qualitative experimental results which show that our algorithm effectively removes the bias information from feature embedding.
《Dynamic Runtime Feature Map Pruning》
arXiv：https://arxiv.org/abs/1812.09922
github：https://github.com/liangtailin/darknet-modified
注：感觉蛮有意思的，代码已经开源
High bandwidth requirements are an obstacle for accelerating the training and inference of deep neural networks. Most previous research focuses on reducing the size of kernel maps for inference. We analyze parameter sparsity of six popular convolutional neural networks - AlexNet, MobileNet, ResNet-50, SqueezeNet, TinyNet, and VGG16. Of the networks considered, those using ReLU (AlexNet, SqueezeNet, VGG16) contain a high percentage of 0-valued parameters and can be statically pruned. Networks with Non-ReLU activation functions in some cases may not contain any 0-valued parameters (ResNet-50, TinyNet). We also investigate runtime feature map usage and find that input feature maps comprise the majority of bandwidth requirements when depth-wise convolution and point-wise convolutions used. We introduce dynamic runtime pruning of feature maps and show that 10% of dynamic feature map execution can be removed without loss of accuracy. We then extend dynamic pruning to allow for values within an epsilon of zero and show a further 5% reduction of feature map loading with a 1% loss of accuracy in top-1.
《Informative Object Annotations: Tell Me Something I Don't Know》
arXiv：https://arxiv.org/abs/1812.10358
Capturing the interesting components of an image is a key aspect of image understanding. When a speaker annotates an image, selecting labels that are informative greatly depends on the prior knowledge of a prospective listener. Motivated by cognitive theories of categorization and communication, we present a new unsupervised approach to model this prior knowledge and quantify the informativeness of a description. Specifically, we compute how knowledge of a label reduces uncertainty over the space of labels and utilize this to rank candidate labels for describing an image. While the full estimation problem is intractable, we describe an efficient algorithm to approximate entropy reduction using a tree-structured graphical model. We evaluate our approach on the open-images dataset using a new evaluation set of 10K ground-truth ratings and find that it achieves ~65% agreement with human raters, largely outperforming other unsupervised baseline approaches.