Search

Capturing long-range dependencies during feature extraction is crucial for video-based person re-identification (re-id) since it would help to tackle many challenging problems such as occlusion and dramatic pose variation. Moreover, capturing subtle differences, such as bags and glasses, is indispensable to distinguish similar pedestrians. In this paper, we propose a novel and efficacious Statistic Attention (SA) block which can capture both the long-range dependencies and subtle differences. SA block leverages high-order statistics of feature maps, which contain both long-range and high-order information. By modeling relations with these statistics, SA block can explicitly capture long-range dependencies with less time complexity. In addition, high-order statistics usually concentrate on details of feature maps and can perceive the subtle differences between pedestrians. In this way, SA block is capable of discriminating pedestrians with subtle differences. Furthermore, this lightweight block can be conveniently inserted into existing deep neural networks at any depth to form Statistic Attention Network (SANet). To evaluate its performance, we conduct extensive experiments on two challenging video re-id datasets, showing that our SANet outperforms the state-of-the-art methods. Furthermore, to show the generalizability of SANet, we evaluate it on three image re-id datasets and two more general image classification datasets, including ImageNet. The source code is available at http://vipl.ict.ac.cn/resources/codes/code/SANet code.zip.

Person reidentification (reID) by convolutional neural network (CNN)-based networks has achieved favorable performance in recent years. However, most of existing CNN-based methods do not take full advantage of spatial–temporal context modeling. In fact, the global spatial–temporal context can greatly clarify local distractions to enhance the target feature representation. To comprehensively leverage the spatial–temporal context information, in this work, we present a novel block, interaction–aggregation-update (IAU), for high-performance person reID. First, the spatial–temporal IAU (STIAU) module is introduced. STIAU jointly incorporates two types of contextual interactions into a CNN framework for target feature learning. Here, the spatial interactions learn to compute the contextual dependencies between different body parts of a single frame, while the temporal interactions are used to capture the contextual dependencies between the same body parts across all frames. Furthermore, a channel IAU (CIAU) module is designed to model the semantic contextual interactions between channel features to enhance the feature representation, especially for small-scale visual cues and body parts. Therefore, the IAU block enables the feature to incorporate the globally spatial, temporal, and channel context. It is lightweight, end-to-end trainable, and can be easily plugged into existing CNNs to form IAUnet. The experiments show that IAUnet performs favorably against state of the art on both image and video reID tasks and achieves compelling results on a general object categorization task. The source code is available at https://github.com/blue-blue272/ImgReID-IAnet .

Background There is a large group of deaf-mutes in the world, and sign language is their major communication tool. Therefore, it’s necessary for the deaf-mutes to communicate with hearing-speech people and the hearing-speech people also have needed to understand sign language, which produces a great demand for sign language teaching. Even though there have already been a large number of books for sign language, it is low efficient to learn sign language with books, even teaching videos. To solve this problem, we develop a smartphone-based interactive Chinese sign language teaching systemfor sign language learning. Methods The system provides a learner with some kinds of learning modes and captures the learner’s actions from its front camera of the smartphone. Right now the system provides a vocabulary set with 1000 frequently used words, and the learner can evaluate his/her sign action by subjective or objective comparison. In the mode of word recognition, the users can play any word within the vocabulary and the system will turn the top three retrieved candidates, so it can remind the learners what the sign is. Results This system provides interactive learning for a user to learn sign language high efficiently. The systemadopts an algorithm based on point cloud recognition to evaluate a user’s sign and costs about 700ms inference time for each sample, which meets the real-time requirements. Conclusion This interactive learning system decreases the communication barriers between the deaf-mutes and hearing-speechers.

In this paper, we address the semantic segmentation task with a new context aggregation scheme named object context, which focuses on enhancing the role of object information. Motivated by the fact that the category of each pixel is inherited from the object it belongs to, we define the object context for each pixel as the set of pixels that belong to the same category as the given pixel in the image. We use a binary relation matrix to represent the relationship between all pixels, where the value one indicates the two selected pixels belong to the same category and zero otherwise. We propose to use a dense relation matrix to serve as a surrogate for the binary relation matrix. The dense relation matrix is capable to emphasize the contribution of object information as the relation scores tend to be larger on the object pixels than the other pixels. Considering that the dense relation matrix estimation requires quadratic computation overhead and memory consumption w.r.t. the input size, we propose an efficient interlaced sparse self-attention scheme to model the dense relations between any two of all pixels via the combination of two sparse relation matrices. To capture richer context information, we further combine our interlaced sparse self-attention scheme with the conventional multi-scale context schemes including pyramid pooling (Zhao et al. 2017) and atrous spatial pyramid pooling (Chen et al. 2018). We empirically show the advantages of our approach with competitive performances on five challenging benchmarks including: Cityscapes, ADE20K, LIP, PASCAL-Context and COCO-Stuff.

State-of-the-art classification models are usually considered as black boxes since their decision processes are implicit to humans. On the contrary, human experts classify objects according to a set of explicit hierarchical criteria. For example, “ tabby is a domestic cat with stripes, dots, or lines”, where tabby is defined by combining its superordinate category ( domestic cat ) and some certain attributes (e.g., has stripes). Inspired by this mechanism, we propose an interpretable Hierarchical Criteria Network (HCN) by additionally learning such criteria. To achieve this goal, images and semantic entities (e.g., taxonomies and attributes) are embedded into a common space, where each category can be represented by the linear combination of its superordinate category and a set of learned discriminative attributes. Specifically, a two-stream convolutional neural network (CNN) is elaborately devised, which embeds images and taxonomies with the two streams respectively. The model is trained by minimizing the prediction error of hierarchy labels on both streams. Extensive experiments on two widely studied datasets (CIFAR-100 and ILSVRC) demonstrate that HCN can learn meaningful attributes as well as reasonable and interpretable classification criteria. Therefore, the proposed method enables further human feedback for model correction as an additional benefit.

Face presentation attack detection (PAD) is essential for securing the widely used face recognition systems. Most of the existing PAD methods do not generalize well to unseen scenarios because labeled training data of the new domain is usually not available. In light of this, we propose an unsupervised domain adaptation with disentangled representation (DR-UDA) approach to improve the generalization capability of PAD into new scenarios. DR-UDA consists of three modules, i.e., ML-Net, UDA-Net and DR-Net. ML-Net aims to learn a discriminative feature representation using the labeled source domain face images via metric learning. UDA-Net performs unsupervised adversarial domain adaptation in order to optimize the source domain and target domain encoders jointly, and obtain a common feature space shared by both domains. As a result, the source domain PAD model can be effectively transferred to the unlabeled target domain for PAD. DR-Net further disentangles the features irrelevant to specific domains by reconstructing the source and target domain face images from the common feature space. Therefore, DR-UDA can learn a disentangled representation space which is generative for face images in both domains and discriminative for live vs. spoof classification. The proposed approach shows promising generalization capability in several public-domain face PAD databases.

Location is an important distinguishing information for instance segmentation. In this paper, we propose a novel model, called Location Sensitive Network (LSNet), for human instance segmentation. LSNet integrates instance-specific location information into one-stage segmentation framework. Specifically, in the segmentation branch, Pose Attention Module (PAM) encodes the location information into the attention regions through coordinates encoding. Based on the location information provided by PAM, the segmentation branch is able to effectively distinguish instances in feature-level. Moreover, we propose a combination operation named Keypoints Sensitive Combination (KSCom) to utilize the location information from multiple sampling points. These sampling points construct the points representation for instances via human keypoints and random points. Human keypoints provide the spatial locations and semantic information of the instances, and random points expand the receptive fields. Based on the points representation for each instance, KSCom effectively reduces the mis-classified pixels. Our method is validated by the experiments on public datasets. LSNet-5 achieves 56.2 mAP at 18.5 FPS on COCOPersons. Besides, the proposed method is significantly superior to its peers in the case of severe occlusion.

Learning the conjunction of multiple visual concepts shows practical significance in various real world applications (e.g. multi-attribute image retrieval and visual relationship detection). In this paper, we propose Concept Conjunction Recurrent Neural Network (C2RNN) to tackle this problem. With our model, visual concepts involved in a conjunction are mapped into the hidden units and combined in a recurrent way to generate the representation of the concept conjunction, which is then used to compute a concept conjunction classifier as the output. We also present an order invariant version of the proposed method based on attention mechanism to learn the tasks without pre-defined concept order. To tackle concept conjunction learning from multiple semantic domains, we introduce a multiplicative framework to learn the joint representation. Experimental results on multi-attribute image retrieval and visual relationship detection show that our method achieves significantly better performance than other related methods on various datasets.

In this work, we study the problem of deep representation learning on a large face dataset with long-tail distribution. Training convolutional neural networks on such dataset with conventional strategy suffers from imbalance problem which results in biased classification boundary, and the few-shot classes lying in tail parts further make the model prone to overfitting. Aiming to learn more discriminative CNN model from long-tail data, we propose a novel aggregate-and-disperse training schema. Firstly, our proposed method aggregates similar classes in tail part to avoid imbalance problem. Based on the aggregated super classes and those original head classes, a model is pre-trained to capture accurate discrimination in head classes as well as coarse discrinimation in tail classes. Secondly, we selectively disperses those aggregated super classes to learn precise inter-class variations and refine the representation for better generalization. We perform extensive experiments on MS-Celeb-1M, BLUFR and MegaFace. Compared with baselines and existing methods, our method achieves better performance of face recognition, demonstrating its effectiveness of handling long-tail distribution.

Functionality is a fundamental attribute of an object which indicates the capability to be used to perform specific actions. It is critical to empower robots the functionality knowledge in discovering appropriate objects for a task e.g. cut cake using knife. Existing research works have focused on understanding object functionality through human-object-interaction from extensively annotated image or video data and are hard to scale up. In this paper, we (1) mine object-functionality knowledge through pattern-based and model-based methods from text, (2) introduce a novel task on physical object-functionality prediction, which consumes an image and an action query to predict whether the object in the image can perform the action, and (3) propose a method to leverage the mined functionality knowledge for the new task. Our experimental results show the effectiveness of our methods.

Due to the highly complex semantic information of images, even with the same query image, the expected content-based image retrieval results could be very different and personalized in different scenarios. However, most existing hashing methods only preserve one single type of semantic similarity, making them incapable of addressing such realistic retrieval tasks. To deal with this problem, we propose a unified hashing framework to encode multiple types of information into the binary codes by exploiting convolutional networks (CNNs). Specifically, we assume that typical retrieval tasks are generally defined in two aspects, i.e. high-level semantics (e.g. object categories) and visual attributes (e.g. object shape and color). To this end, our Dual Purpose Hashing model is trained to jointly preserve two kinds of similarities characterizing the two aspects respectively. Moreover, since images with both category and attribute labels are scarce, our model is carefully designed to leverage the abundant partially labelled data as training inputs to alleviate the risk of overfitting. With such a framework, the binary codes of new-coming images can be readily obtained by quantizing the outputs of a specific CNN layer, and different retrieval tasks can be achieved by using the binary codes in different ways. Experiments on two large-scale datasets show that our method achieves comparable or even better performance than those state-of-the-art methods specifically designed for each individual retrieval task while being more compact than the compared methods.

Solving visual question answering (VQA) task requires recognizing many diverse visual concepts as the answer. These visual concepts contain rich structural semantic meanings, e.g., some concepts in VQA are highly related (e.g., red & blue), some of them are less relevant (e.g., red & standing). It is very natural for humans to efficiently learn concepts by utilizing their semantic meanings to concentrate on distinguishing relevant concepts and eliminate the disturbance of irrelevant concepts. However, previous works usually use a simple MLP to output visual concept as the answer in a flat label space that treats all labels equally, causing limitations in representing and using the semantic meanings of labels. To address this issue, we propose a novel visual recognition module named Dynamic Concept Recognizer (DCR), which is easy to be plugged in an attention-based VQA model, to utilize the semantics of the labels in answer prediction. Concretely, we introduce two key features in DCR: 1) a novel structural label space to depict the difference of semantics between concepts, where the labels in new label space are assigned to different groups according to their meanings. This type of semantic information helps decompose the visual recognizer in VQA into multiple specialized sub-recognizers to improve the capacity and efficiency of the recognizer. 2) A feature attention mechanism to capture the similarity between relevant groups of concepts, e.g., human-related group “chef, waiter” is more related to “swimming, running, etc.” than scene related group “sunny, rainy, etc.”. This type of semantic information helps sub-recognizers for relevant groups to adaptively share part of modules and to share the knowledge between relevant sub-recognizers to facilitate the learning procedure. Extensive experiments on several datasets have shown that the proposed structural label space and DCR module can efficiently learn the visual concept recognition and benefit the performance of the VQA model.

We introduce a new efficient framework, the Unified Context Network (UniCon), for robust active speaker detection (ASD). Traditional methods for ASD usually operate on each candidate's pre-cropped face track separately and do not sufficiently consider the relationships among the candidates. This potentially limits performance, especially in challenging scenarios with low-resolution faces, multiple candidates, etc. Our solution is a novel, unified framework that focuses on jointly modeling multiple types of contextual information: spatial context to indicate the position and scale of each candidate's face, relational context to capture the visual relationships among the candidates and contrast audio-visual affinities with each other, and temporal context to aggregate long-term information and smooth out local uncertainties. Based on such information, our model optimizes all candidates in a unified process for robust and reliable ASD. A thorough ablation study is performed on several challenging ASD benchmarks under different settings. In particular, our method outperforms the state-of-the-art by a large margin of about 15% mean Average Precision (mAP) absolute on two challenging subsets: one with three candidate speakers, and the other with faces smaller than 64 pixels. Together, our UniCon achieves 92.0% mAP on the AVA-ActiveSpeaker validation set, surpassing 90% for the first time on this challenging dataset at the time of submission. Project website: https://unicon-asd.github.io/., Comment: 10 pages, 6 figures; to appear at ACM Multimedia 2021

Objective The profession of surgery is entering a new era of “big data,” where analyses of longitudinal trainee assessment data will be used to inform ongoing efforts to improve surgical education. Given the high-stakes implications of these types of analyses, researchers must define the conditions under which estimates derived from these large datasets remain valid. With this study, we determine the number of assessments of residents’ performances needed to reliably assess the difficulty of “Core” surgical procedures. Design Using the SIMPL smartphone application from the Procedural Learning and Safety Collaborative, 402 attending surgeons directly observed and provided workplace-based assessments for 488 categorical residents after 5259 performances of 87 Core surgical procedures performed at 14 institutions. We used these faculty ratings to construct a linear mixed model with resident performance as the outcome variable and multiple predictors including, most significantly, the operative procedure as a random effect. We interpreted the variance in performance ratings attributable to the procedure, after controlling for other variables, as the “difficulty” of performing the procedure. We conducted a generalizability analysis and decision study to estimate the number of SIMPL performance ratings needed to reliably estimate the difficulty of a typical Core procedure. Results Twenty-four faculty ratings of resident operative performance were necessary to reliably estimate the difficulty of a typical Core surgical procedure (mean dependability coefficient 0.80, 95% confidence interval 0.73-0.87). Conclusions At least 24 operative performance ratings are required to reliably estimate the difficulty of a typical Core surgical procedure. Future research using performance ratings to establish procedure difficulty should include adequate numbers of ratings given the high-stakes implications of those results for curriculum design and policy.

Facial attribute editing aims to manipulate single or multiple attributes of a face image, i.e., to generate a new face with desired attributes while preserving other details. Recently, generative adversarial net (GAN) and encoder-decoder architecture are usually incorporated to handle this task with promising results. Based on the encoder-decoder architecture, facial attribute editing is achieved by decoding the latent representation of the given face conditioned on the desired attributes. Some existing methods attempt to establish an attribute-independent latent representation for further attribute editing. However, such attribute-independent constraint on the latent representation is excessive because it restricts the capacity of the latent representation and may result in information loss, leading to over-smooth and distorted generation. Instead of imposing constraints on the latent representation, in this work we apply an attribute classification constraint to the generated image to just guarantee the correct change of desired attributes, i.e., to "change what you want". Meanwhile, the reconstruction learning is introduced to preserve attribute-excluding details, in other words, to "only change what you want". Besides, the adversarial learning is employed for visually realistic editing. These three components cooperate with each other forming an effective framework for high quality facial attribute editing, referred as AttGAN. Furthermore, our method is also directly applicable for attribute intensity control and can be naturally extended for attribute style manipulation. Experiments on CelebA dataset show that our method outperforms the state-of-the-arts on realistic attribute editing with facial details well preserved., Submitted to IEEE Transactions on Image Processing, Code: https://github.com/LynnHo/AttGAN-Tensorflow

Visual sign language recognition is an interesting and challenging problem. To create a discriminative representation, a hierarchical Grassmann covariance matrix (HGCM) model is proposed for sign description. Furthermore, a multi-temporal belief propagation (MTBP) based segmentation approach is presented for continuous sequence spotting. Concretely speaking, a sign is represented by multiple covariance matrices, followed by evaluating and selecting their most significant singular vectors. These covariance matrices are transformed into a more compact and discriminative HGCM, which is formulated on the Grassmann manifold. Continuous sign sequences can be recognized frame by frame using the HGCM model, before being optimized by MTBP, which is a carefully designed graphic model. The proposed method is thoroughly evaluated on isolated and synthetic and real continuous sign datasets as well as on HDM05. Extensive experimental results convincingly show the effectiveness of our proposed framework.

The explosive growth of digital images in video surveillance and social media has led to the significant need for efficient search of persons of interest in law enforcement and forensic applications. Despite tremendous progress in primary biometric traits (e.g., face and fingerprint) based person identification, a single biometric trait alone cannot meet the desired recognition accuracy in forensic scenarios. Tattoos, as one of the important soft biometric traits, have been found to be valuable for assisting in person identification. However, tattoo search in a large collection of unconstrained images remains a difficult problem, and existing tattoo search methods mainly focus on matching cropped tattoos, which is different from real application scenarios. To close the gap, we propose an efficient tattoo search approach that is able to learn tattoo detection and compact representation jointly in a single convolutional neural network (CNN) via multi-task learning. While the features in the backbone network are shared by both tattoo detection and compact representation learning, individual latent layers of each sub-network optimize the shared features toward the detection and feature learning tasks, respectively. We resolve the small batch size issue inside the joint tattoo detection and compact representation learning network via random image stitch and preceding feature buffering. We evaluate the proposed tattoo search system using multiple public-domain tattoo benchmarks, and a gallery set with about 300K distracter tattoo images compiled from these datasets and images from the Internet. In addition, we also introduce a tattoo sketch dataset containing 300 tattoos for sketch-based tattoo search. Experimental results show that the proposed approach has superior performance in tattoo detection and tattoo search at scale compared to several state-of-the-art tattoo retrieval algorithms., Comment: Technical Report (15 pages, 14 figures)

Zero-shot learning (ZSL) has enjoyed great popularity in recent years due to its ability to recognize novel objects, where semantic information is exploited to build up relations among different categories. Traditional ZSL approaches usually focus on learning more robust visual-semantic embeddings among seen classes and directly apply them to the unseen classes without considering whether they are suitable. It is well known that domain gap exists between seen and unseen classes. In order to tackle such problem, we propose a novel adaptive metric learning approach to measure the compatibility between visual samples and class semantics, where class similarities are utilized to adapt the visual-semantic embedding to the unseen classes. Extensive experiments on four benchmark ZSL datasets show the effectiveness of the proposed approach.

Attributes are mid-level semantic properties of objects. Recent research has shown that visual attributes can benefit many typical learning problems in computer vision community. However, attribute learning is still a challenging problem as the attributes may not always be predictable directly from input images and the variation of visual attributes is sometimes large across categories. In this paper, we propose a unified multiplicative framework for attribute learning, which tackles the key problems. Specifically, images and category information are jointly projected into a shared feature space, where the latent factors are disentangled and multiplied to fulfil attribute prediction. The resulting attribute classifier is category-specific instead of being shared by all categories. Moreover, our model can leverage auxiliary data to enhance the predictive ability of attribute classifiers, which can reduce the effort of instance-level attribute annotation to some extent. By integrated into an existing deep learning framework, our model can both accurately predict attributes and learn efficient image representations. Experimental results show that our method achieves superior performance on both instance-level and category-level attribute prediction. For zero-shot learning based on visual attributes and human-object interaction recognition, our method can improve the state-of-the-art performance on several widely used datasets.

Although the conventional active appearance model (AAM) has achieved some success for face alignment, it still suffers from the generalization problem when be applied to unseen subjects and images. To deal with the generalization problem of AAM, we first reformulate the original AAM as sparsity-regularized AAM, which can achieve more compact/better shape and appearance priors by selecting nearest neighbors as the bases of the shape and appearance model. To speed up the fitting procedure, the sparsity in sparsity-regularized AAM is approximated by using the locality (i.e., K-nearest neighbor), and thus inducing the locality-constrained active appearance-model (LC-AAM). The LC-AAM solves a constrained AAM-like fitting problem with the K-nearest neighbors as the bases of shape and appearance model. To alleviate the adverse influence of inaccurate K-nearest neighbor results, the locality constraint is further embedded in the discriminative fitting method denoted as LC-DFM, which can find better K-nearest neighbor results by employing shape-indexed feature, and can also tolerate some inaccurate neighbors benefited from the regression model rather than the generative model in AAM. Extensive experiments on several datasets demonstrate that our methods outperform the state-of-the-arts in both detection accuracy and generalization ability.

In this paper, we present a new hashing method to learn compact binary codes for highly efficient image retrieval on large-scale datasets. While the complex image appearance variations still pose a great challenge to reliable retrieval, in light of the recent progress of Convolutional Neural Networks (CNNs) in learning robust image representation on various vision tasks, this paper proposes a novel Deep Supervised Hashing method to learn compact similarity-preserving binary code for the huge body of image data. Specifically, we devise a CNN architecture that takes pairs/triplets of images as training inputs and encourages the output of each image to approximate discrete values (e.g. $$+\,1$$/$$-\,1$$). To this end, the loss functions are elaborately designed to maximize the discriminability of the output space by encoding the supervised information from the input image pairs/triplets, and simultaneously imposing regularization on the real-valued outputs to approximate the desired discrete values. For image retrieval, new-coming query images can be easily encoded by forward propagating through the network and then quantizing the network outputs to binary codes representation. Extensive experiments on three large scale datasets CIFAR-10, NUS-WIDE, and SVHN show the promising performance of our method compared with the state-of-the-arts.

Expressive representations for characterizing face appearances are essential for accurate face detection. Due to different poses, scales, illumination, occlusion, etc, face appearances generally exhibit substantial variations, and the contents of each local region (facial part) vary from one face to another. Current detectors, however, particularly those based on convolutional neural networks, apply identical operations (e.g. convolution or pooling) to all local regions on each face for feature aggregation (in a generic sliding-window configuration), and take all local features as equally effective for the detection task. In such methods, not only is each local feature suboptimal due to ignoring region-wise distinctions, but also the overall face representations are semantically inconsistent. To address the issue, we design a hierarchical attention mechanism to allow adaptive exploration of local features. Given a face proposal, part-specific attention modeled as learnable Gaussian kernels is proposed to search for proper positions and scales of local regions to extract consistent and informative features of facial parts. Then face-specific attention predicted with LSTM is introduced to model relations between the local parts and adjust their contributions to the detection tasks. Such hierarchical attention leads to a part-aware face detector, which forms more expressive and semantically consistent face representations. Extensive experiments are performed on three challenging face detection datasets to demonstrate the effectiveness of our hierarchical attention and make comparisons with state-of-the-art methods.

Image caption evaluation is a crucial task, which involves the semantic perception and matching of image and text. Good evaluation metrics aim to be fair, comprehensive, and consistent with human judge intentions. When humans evaluate a caption, they usually consider multiple aspects, such as whether it is related to the target image without distortion, how much image gist it conveys, as well as how fluent and beautiful the language and wording is. The above three different evaluation orientations can be summarized as fidelity, adequacy, and fluency. The former two rely on the image content, while fluency is purely related to linguistics and more subjective. Inspired by human judges, we propose a learning-based metric named FAIEr to ensure evaluating the fidelity and adequacy of the captions. Since image captioning involves two different modalities, we employ the scene graph as a bridge between them to represent both images and captions. FAIEr mainly regards the visual scene graph as the criterion to measure the fidelity. Then for evaluating the adequacy of the candidate caption, it high-lights the image gist on the visual scene graph under the guidance of the reference captions. Comprehensive experimental results show that FAIEr has high consistency with human judgment as well as high stability, low reference dependency, and the capability of reference-free evaluation.

Catastrophic forgetting, the main challenge of Class Incremental Learning, is closely related to the classifier’s bias due to imbalanced data, and most researchers resort to empirical techniques to remove the bias. Such anti-bias tricks share many ideas with the field of Class Imbalanced Learning, which encourages us to reflect on why these tricks work, and how we can design more principled solutions from a different perspective. In this paper, we comprehensively analyze the connections and seek possible collaborations between these two fields, i.e. Class Incremental Learning and Class Imbalanced Learning. Specifically, we first provide a panoramic view of recent bias correction tricks from the perspective of handling class imbalance. Then, we show that an adapted post-scaling technique which originates from Class Imbalanced Learning is on par with or even outperforms SOTA Class Incremental Learning method. Visualization via violin plots and polar charts further sheds light on how SOTA methods address the class imbalance problem from a more intuitive geometric perspective. These findings may encourage further infiltration between the two closely connected fields, but also raise concerns about whether it is correct that Class Incremental Learning degenerates into a class imbalance problem.

Person re-identification (reID) plays an important role in computer vision. However, existing methods suffer from performance degradation in occluded scenes. In this work, we propose an occlusion-robust block, Region Feature Completion (RFC), for occluded reID. Different from most previous works that discard the occluded regions, RFC block can recover the semantics of occluded regions in feature space. Firstly, a Spatial RFC (SRFC) module is developed. SRFC exploits the long-range spatial contexts from non-occluded regions to predict the features of occluded regions. The unit-wise prediction task leads to an encoder/decoder architecture, where the region-encoder models the correlation between non-occluded and occluded region, and the region-decoder utilizes the spatial correlation to recover occluded region features. Secondly, we introduce Temporal RFC (TRFC) module which captures the long-term temporal contexts to refine the prediction of SRFC. RFC block is lightweight, end-to-end trainable and can be easily plugged into existing CNNs to form RFCnet. Extensive experiments are conducted on occluded and commonly holistic reID benchmarks. Our method significantly outperforms existing methods on the occlusion datasets, while remains top even superior performance on holistic datasets. The source code is available at https://github.com/blue-blue272/OccludedReID-RFCnet., 18 pages, 17 figures. The paper is accepted by TPAMI, and the code is available at https://github.com/blue-blue272/OccludedReID-RFCnet

Sentence is not only the basic unit of natural language, but also the research object of Nature Language Process. Sentence similarity computing is the basic of text similarity computing, so sentence similarity evaluation set is an essential data set for similarity technology research. Only by establishing an appropriate evaluation set can we objectively evaluate the advantages and disadvantages of similarity computing methods. In order to objectively evaluate the performance of Tibetan sentence similarity, this paper designs the construction scheme of Tibetan sentence similarity evaluation set, and builds the evaluation set TSS_320 for evaluating Tibetan sentence similarity, based on the analysis of the construction method of English and Chinese sentence similarity evaluation set, also combined with the characteristics of Tibetan sentences. The validity of the evaluation set is verified by statistical methods.

Sentence similarity calculation has an important position in natural language processing and is the basis of text similarity calculation, therefore its accuracy affects the performance of natural language processing. In order to solve the problem of Tibetan sentence similarity calculation, this paper analyzes the contribution of sentence edit distance, sentence structure, sentence length, word order, and the same word to sentence similarity, designs a multi-feature fusion model for Tibetan sentence similarity calculation, proposes a multi-feature fusion method for Tibetan sentence similarity calculation. Experimental results show that the proposed method has achieved good results in Tibetan sentence similarity calculations.

Dense video event captioning aims to generate a sequence of descriptive captions for each event in a long untrimmed video. Video-level context provides important information and facilities the model to generate consistent and less redundant captions between events. In this paper, we introduce a novel Hierarchical Context-aware Network for dense video event captioning (HCN) to capture context from various aspects. In detail, the model leverages local and global context with different mechanisms to jointly learn to generate coherent captions. The local context module performs full interaction between neighbor frames and the global context module selectively attends to previous or future events. According to our extensive experiment on both Youcook2 and Activitynet Captioning datasets, the video-level HCN model outperforms the event-level context-agnostic model by a large margin. The code is available at https://github.com/KirkGuo/HCN.

For a given text, previous text-to-image synthesis methods commonly utilize a multistage generation model to produce images with high resolution in a coarse-to-fine manner. However, these methods ignore the interaction among stages, and they do not constrain the consistent cross-sample relations of images generated in different stages. These deficiencies result in inefficient generation and discrimination. In this study, we propose an interstage cross-sample similarity distillation model based on a generative adversarial network (GAN) for learning efficient text-to-image synthesis. To strengthen the interaction among stages, we achieve interstage knowledge distillation from the refined stage to the coarse stages with novel interstage cross-sample similarity distillation blocks. To enhance the constraint on the cross-sample relations of the images generated at different stages, we conduct cross-sample similarity distillation among the stages. Extensive experiments on the Oxford-102 and Caltech-UCSD Birds-200-2011 (CUB) datasets show that our model generates visually pleasing images and achieves quantitatively comparable performance with state-of-the-art methods.

Video face retrieval (VFR) is an appealing and practical computer vision task, which aims to search particular character from masses of videos like in TV-Series. The challenges of this task mainly lie in two aspects, i.e. faces in such videos contain complex appearance variations with uncontrollable shooting environment and searching from big data usually requires high efficiency in both space and time. To fulfill this task, current works typically proceed in a learning to hash manner by fusing single-frame features within a video to obtain the video representation and further embedding it into Hamming space to yield video binary codes. The feature fusion stage has inevitably discarded too much frame information and leads to less discriminative video codes. In this paper, we propose Hybrid Video and Image Hashing (HVIH) to learn more effective binary codes for face videos. Specifically, we fully exploit the dense frame features rather than simply discarding them after the video level fusion and jointly optimize binary codes for the video and its composed frames in adapted supervised manners. To achieve more robust video representation, we introduce a module of video center alignment to ensure the binary codes location of the video and its frames to be as compact and consistent as possible in the Hamming space, which naturally facilitates both tasks of video-to-video retrieval and image-to-video retrieval. Extensive experiments on two challenging video face databases demonstrate the superiority of our approach over the state-of-the-art.

Human poses play important roles in action analysis. However, most state-of-the-art approaches in action recognition ignore the importance of human poses and rarely leverage the pose information for further improving the recognition performance. In this paper, we propose a novel network architecture, which simultaneously considers the appearance information and pose knowledge for robust action recognition. We explore various architectures for fusing the appearance and pose information rather than simply averaging scores at the final layer. Moreover, a novel training strategy is proposed to reduce the influence of overfitting for limited training data. Extensive experiments show that our method achieves competitive performance on the popular benchmarks, i.e., UCF-101 and HMDB-51.

Video captioning is a fundamental task for visual understanding. Previous works employ end-to-end networks to learn from the low-level vision feature and generate descriptive captions, which are hard to recognize fine-grained objects and lacks the understanding of crucial semantic concepts. According to DPC [19], these concepts generally present in the narrative transcripts of the instructional videos. The incorporation of transcript and video can improve the captioning performance. However, DPC directly concatenates the embedding of transcript with video features, which is incapable of fusing language and vision features effectively and leads to the temporal mis-alignment between transcript and video. This motivates us to 1) learn the semantic concepts explicitly and 2) design a temporal alignment mechanism to better align the video and transcript for the captioning task. In this paper, we start with an encoder-decoder backbone using transformer models. Firstly, we design a semantic concept prediction module as a multi-task to train the encoder in a supervised way. Then, we develop an attention based cross-modality temporal alignment method that combines the sequential video frames and transcript sentences. Finally, we adopt a copy mechanism to enable the decoder(generation) module to copy important concepts from source transcript directly. The extensive experimental results demonstrate the effectiveness of our model, which achieves state-of-the-art results on YouCookII dataset.

3D models are commonly used in computer vision and graphics. With the wider availability of mesh data, an efficient and intrinsic deep learning approach to processing 3D meshes is in great need. Unlike images, 3D meshes have irregular connectivity, requiring careful design to capture relations in the data. To utilize the topology information while staying robust under different triangulations, we propose to encode mesh connectivity using Laplacian spectral analysis, along with mesh feature aggregation blocks (MFABs) that can split the surface domain into local pooling patches and aggregate global information amongst them. We build a mesh hierarchy from fine to coarse using Laplacian spectral clustering, which is flexible under isometric transformations. Inside the MFABs there are pooling layers to collect local information and multi-layer perceptrons to compute vertex features of increasing complexity. To obtain the relationships among different clusters, we introduce a Correlation Net to compute a correlation matrix, which can aggregate the features globally by matrix multiplication with cluster features. Our network architecture is flexible enough to be used on meshes with different numbers of vertices. We conduct several experiments including shape segmentation and classification, and our method outperforms state-of-the-art algorithms for these tasks on the ShapeNet and COSEG datasets.

The state of the art person search methods separate person search into detection and re-ID stages, but ignore the consistency between these two stages. The general person detector has no special attention on the query target; The re-ID model is trained on hand-drawn bounding boxes which are not available in person search. To address the consistency problem, we introduce a Task-Consist Two-Stage (TCTS) person search framework, includes an identity-guided query (IDGQ) detector and a Detection Results Adapted (DRA) re-ID model. In the detection stage, the IDGQ detector learns an auxiliary identity branch to compute query similarity scores for proposals. With consideration of the query similarity scores and foreground score, IDGQ produces query-like bounding boxes for the re-ID stage. In the re-ID stage, we predict identity labels of detected bounding boxes, and use these examples to construct a more practical mixed train set for the DRA model. Training on the mixed train set improves the robustness of the re-ID stage to inaccurate detection. We evaluate our method on two benchmark datasets, CUHK-SYSU and PRW. Our framework achieves 93.9% of mAP and 95.1% of rank1 accuracy on CUHK-SYSU, outperforming the previous state of the art methods.

Answering questions that require reading texts in an image is challenging for current models. One key difficulty of this task is that rare, polysemous, and ambiguous words frequently appear in images, e.g., names of places, products, and sports teams. To overcome this difficulty, only resorting to pre-trained word embedding models is far from enough. A desired model should utilize the rich information in multiple modalities of the image to help understand the meaning of scene texts, e.g., the prominent text on a bottle is most likely to be the brand. Following this idea, we propose a novel VQA approach, Multi-Modal Graph Neural Network (MM-GNN). It first represents an image as a graph consisting of three sub-graphs, depicting visual, semantic, and numeric modalities respectively. Then, we introduce three aggregators which guide the message passing from one graph to another to utilize the contexts in various modalities, so as to refine the features of nodes. The updated nodes have better features for the downstream question answering module. Experimental evaluations show that our MM-GNN represents the scene texts better and obviously facilitates the performances on two VQA tasks that require reading scene texts., Published as a CVPR2020 paper

Point clouds contain rich spatial information, which provides complementary cues for gesture recognition. In this paper, we formulate gesture recognition as an irregular sequence recognition problem and aim to capture long-term spatial correlations across point cloud sequences. A novel and effective PointLSTM is proposed to propagate information from past to future while preserving the spatial structure. The proposed PointLSTM combines state information from neighboring points in the past with current features to update the current states by a weight-shared LSTM layer. This method can be integrated into many other sequence learning approaches. In the task of gesture recognition, the proposed PointLSTM achieves state-of-the-art results on two challenging datasets (NVGesture and SHREC'17) and outperforms previous skeleton-based methods. To show its advantages in generalization, we evaluate our method on MSR Action3D dataset, and it produces competitive results with previous skeleton-based methods.

Image-level weakly supervised semantic segmentation is a challenging problem that has been deeply studied in recent years. Most of advanced solutions exploit class activation map (CAM). However, CAMs can hardly serve as the object mask due to the gap between full and weak supervisions. In this paper, we propose a self-supervised equivariant attention mechanism (SEAM) to discover additional supervision and narrow the gap. Our method is based on the observation that equivariance is an implicit constraint in fully supervised semantic segmentation, whose pixel-level labels take the same spatial transformation as the input images during data augmentation. However, this constraint is lost on the CAMs trained by image-level supervision. Therefore, we propose consistency regularization on predicted CAMs from various transformed images to provide self-supervision for network learning. Moreover, we propose a pixel correlation module (PCM), which exploits context appearance information and refines the prediction of current pixel by its similar neighbors, leading to further improvement on CAMs consistency. Extensive experiments on PASCAL VOC 2012 dataset demonstrate our method outperforms state-of-the-art methods using the same level of supervision. The code is released online., To appear at CVPR 2020 (Oral)

Domain adaptation attempts to boost the performance on a target domain by borrowing knowledge from a well established source domain. To handle the distribution gap between two domains, the prominent approaches endeavor to extract domain-invariant features. It is known that after a perfect domain alignment the domain-invariant representations of two domains should share the same characteristics from perspective of the overview and also any local piece. Inspired by this, we propose a novel method called Hierarchical Gradient Synchronization to model the synchronization relationship among the local distribution pieces and global distribution, aiming for more precise domain-invariant features. Specifically, the hierarchical domain alignments including class-wise alignment, group-wise alignment and global alignment are first constructed. Then, these three types of alignment are constrained to be consistent to ensure better structure preservation. As a result, the obtained features are domain invariant and intrinsically structure preserved. As evaluated on extensive domain adaptation tasks, our proposed method achieves state-of-the-art classification performance on both vanilla unsupervised domain adaptation and partial domain adaptation.

Inspired by the concept of memory mechanism and predictive coding from the cognitive neuroscience, this paper presents a deep memory and prediction neural network (DMPNet) for video prediction. Correspondingly, memory and error propagation units are designed in DMPNet to capture the previous spatial-temporal information and compute current predictive error which is forwarded to the prediction unit for correcting the subsequent video prediction. Subsequently, prediction unit takes the information stored in memory unit and predictive error of previous frame as input to predict the next frame. We evaluate our method on two public real-world datasets and demonstrate that the proposed DMPNet outperforms some state-of-the-art methods quantitatively and qualitatively.

Lip reading has received an increasing research interest in recent years due to the rapid development of deep learning and its widespread potential applications. One key point to obtain good performance for the lip reading task depends heavily on how effective the representation can be to capture the lip movement information and meanwhile to resist the noises resulted from the change of pose, lighting conditions, speaker's appearance and so on. Towards this target, we propose to introduce the mutual information constraints on both the local feature's level and the global sequence's level to enhance the relations of the features with the speech content. On the one hand, we constraint the features generated at each time step to enable them carry a strong relation with the speech content by imposing the local mutual information maximization constraint (LMIM), leading to improvements over the model's ability to discover fine-grained lip movements and the fine-grained differences among words with similar pronunciation, such as ``spend'' and ``spending''. On the other hand, we introduce the mutual information maximization constraint on the global sequence's level (GMIM), to make the model be able to pay more attention to discriminate key frames related with the speech content, and less to various noises appeared in the speaking process. By combining these two advantages together, the proposed method is expected to be both discriminative and robust for effective lip reading. To verify this method, we evaluate on two large-scale benchmark. We perform a detailed analysis and comparison on several aspects, including the comparison of the LMIM and GMIM with the baseline, the visualization of the learned representation and so on. The results not only prove the effectiveness of the proposed method but also report new state-of-the-art performance on both the two benchmarks., 8 pages, Accepted in the 15th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2020)

Preserving the semantic similarity is one of the most important goals of hashing. Most existing deep hashing methods employ pairs or triplets of samples in training stage, which only consider the semantic similarity within a minibatch and depict the local positional relationship in Hamming space, leading to intermittent semantic similarity preservation. In this paper, we propose Deep Position-Aware Hashing (DPAH) to ensure continuous semantic similarity in Hamming space by modeling global positional relationship. Specifically, we introduce a set of learnable class centers as the global proxies to represent the global information and generate discriminative binary codes by constraining the distance between data points and class centers. In addition, in order to reduce the information loss caused by relaxing the binary codes to real-values in optimization, we propose kurtosis loss (KT loss) to handle the distribution of real-valued features before thresholding to be double-peak, and then enable the real-valued features to be more binarylike. Comprehensive experiments on three datasets show that our DPAH outperforms state-of-the-art methods.

Face attribute estimation has many potential applications in video surveillance, face retrieval, and social media. While a number of methods have been proposed for face attribute estimation, most of them did not explicitly consider the attribute correlation and heterogeneity (e.g., ordinal vs. nominal and holistic vs. local) during feature representation learning. In this paper, we present a Deep Multi-Task Learning (DMTL) approach to jointly estimate multiple heterogeneous attributes from a single face image. In DMTL, we tackle attribute correlation and heterogeneity with convolutional neural networks (CNNs) consisting of shared feature learning for all the attributes, and category-specific feature learning for heterogeneous attributes. We also introduce an unconstrained face database (LFW+), an extension of public-domain LFW, with heterogeneous demographic attributes (age, gender, and race) obtained via crowdsourcing. Experimental results on benchmarks with multiple face attributes (MORPH II, LFW+, CelebA, LFWA, and FotW) show that the proposed approach has superior performance compared to state of the art. Finally, evaluations on a public-domain face database (LAP) with a single attribute show that the proposed approach has excellent generalization ability., Comment: To appear in the IEEE Trans. Pattern Analysis and Machine Intelligence (final)

Attributes are widely used in different vision tasks. However, existing attribute resources are quite limited and most of them are not in large scale. Current attribute annotation process is generally done by human, which is expensive and time-consuming. In this paper, we propose a novel framework to perform effective attribute annotations. Based on the common knowledge that attributes can be shared among different classes, we leverage the benefits of transfer learning and active learning together to transfer knowledge from some existing small attribute databases to large-scale target databases. In order to learn more robust attribute models, attribute relationships are incorporated to assist the learning process. Using the proposed framework, we conduct extensive experiments on two large-scale image databases, i.e. ImageNet and SUN Attribute, where high quality automatic attribute annotations are obtained.

High accuracy face recognition is of great importance for a wide variety of real-world applications. Although significant progress has been made in the last decades, fully automatic face recognition systems have not yet approached the goal of surpassing the human vision system, even in controlled conditions. In this paper, we propose an approach for robust face recognition by fusing two complementary features: one is Gabor magnitude of multiple scales and orientations and the other is Fourier phase encoded by spatial pyramid based local phase quantization (SPLPQ). To reduce the high dimensionality of both features, block-wise fisher discriminant analysis (BFDA) is applied and further combined by score-level fusion. Moreover, inspired by the biological cognitive mechanism, multiple face models are exploited to further boost the robustness of the proposed approach. We evaluate the proposed approach on three challenging databases, i.e., FRGC ver2.0, LFW, and CFW-p, that address two face classification scenarios, i.e., verification and identification. Experimental results consistently exhibit the complementarity of the two features and the performance boost gained by the multiple face models. The proposed approach achieved approximately 96% verification rate when FAR was 0.1% on FRGC ver2.0 Exp.4, impressively surpassing all the best known results.

Due to the imperfect person detection results and posture changes, temporal appearance misalignment is unavoidable in video-based person re-identification (ReID). In this case, 3D convolution may destroy the appearance representation of person video clips, thus it is harmful to ReID. To address this problem, we propose Appearance-Preserving 3D Convolution (AP3D), which is composed of two components: an Appearance-Preserving Module (APM) and a 3D convolution kernel. With APM aligning the adjacent feature maps in pixel level, the following 3D convolution can model temporal information on the premise of maintaining the appearance representation quality. It is easy to combine AP3D with existing 3D ConvNets by simply replacing the original 3D convolution kernels with AP3Ds. Extensive experiments demonstrate the effectiveness of AP3D for video-based ReID and the results on three widely used datasets surpass the state-of-the-arts. Code is available at: https://github.com/guxinqian/AP3D.

Although two-stage object detectors have continuously advanced the state-of-the-art performance in recent years, the training process itself is far from crystal. In this work, we first point out the inconsistency problem between the fixed network settings and the dynamic training procedure, which greatly affects the performance. For example, the fixed label assignment strategy and regression loss function cannot fit the distribution change of proposals and thus are harmful to training high quality detectors. Consequently, we propose Dynamic R-CNN to adjust the label assignment criteria (IoU threshold) and the shape of regression loss function (parameters of SmoothL1 Loss) automatically based on the statistics of proposals during training. This dynamic design makes better use of the training samples and pushes the detector to fit more high quality samples. Specifically, our method improves upon ResNet-50-FPN baseline with 1.9% AP and 5.5% AP\(_{90}\) on the MS COCO dataset with no extra overhead. Codes and models are available at https://github.com/hkzhang95/DynamicRCNN.

Scene graph aims to faithfully reveal humans’ perception of image content. When humans analyze a scene, they usually prefer to describe image gist first, namely major objects and key relations in a scene graph. This humans’ inherent perceptive habit implies that there exists a hierarchical structure about humans’ preference during the scene parsing procedure. Therefore, we argue that a desirable scene graph should be also hierarchically constructed, and introduce a new scheme for modeling scene graph. Concretely, a scene is represented by a human-mimetic Hierarchical Entity Tree (HET) consisting of a series of image regions. To generate a scene graph based on HET, we parse HET with a Hybrid Long Short-Term Memory (Hybrid-LSTM) which specifically encodes hierarchy and siblings context to capture the structured information embedded in HET. To further prioritize key relations in the scene graph, we devise a Relation Ranking Module (RRM) to dynamically adjust their rankings by learning to capture humans’ subjective perceptive habits from objective entity saliency and size. Experiments indicate that our method not only achieves state-of-the-art performances for scene graph generation, but also is expert in mining image-specific relations which play a great role in serving downstream tasks.

This paper proposes a Temporal Complementary Learning Network that extracts complementary features of consecutive video frames for video person re-identification. Firstly, we introduce a Temporal Saliency Erasing (TSE) module including a saliency erasing operation and a series of ordered learners. Specifically, for a specific frame of a video, the saliency erasing operation drives the specific learner to mine new and complementary parts by erasing the parts activated by previous frames. Such that the diverse visual features can be discovered for consecutive frames and finally form an integral characteristic of the target identity. Furthermore, a Temporal Saliency Boosting (TSB) module is designed to propagate the salient information among video frames to enhance the salient feature. It is complementary to TSE by effectively alleviating the information loss caused by the erasing operation of TSE. Extensive experiments show our method performs favorably against state-of-the-arts. The source code is available at https://github.com/blue-blue272/VideoReID-TCLNet.

We present a model-agnostic post-processing scheme to improve the boundary quality for the segmentation result that is generated by any existing segmentation model. Motivated by the empirical observation that the label predictions of interior pixels are more reliable, we propose to replace the originally unreliable predictions of boundary pixels by the predictions of interior pixels. Our approach processes only the input image through two steps: (i) localize the boundary pixels and (ii) identify the corresponding interior pixel for each boundary pixel. We build the correspondence by learning a direction away from the boundary pixel to an interior pixel. Our method requires no prior information of the segmentation models and achieves nearly real-time speed. We empirically verify that our SegFix consistently reduces the boundary errors for segmentation results generated from various state-of-the-art models on Cityscapes, ADE20K and GTA5. Code is available at: https://github.com/openseg-group/openseg.pytorch.

Unsupervised domain adaptation aims to train a good model for a target domain via transferring knowledge from a related labeled source domain, thus reducing the dependency on huge labeling of target domain samples. Generative adversarial net (GAN) is a newly proposed technique which has shown its capability of alleviating distribution discrepancy. Inspired by GAN, in this work, we propose a novel duplex GAN (DupGAN) which extracts domain invariant and discriminative representation guided by bidirectional domain transformation, formulated as a GAN with duplex discriminators. In addition, each of the duplex discriminators not only judges reality/falsity, but also performs category classification for real images to preserve the category information during domain transformation. As evaluated on the standard benchmarks, i.e., digits datasets and Office-31, our proposed DupGAN outperforms the state-of-the-art methods, indicating its effectiveness on unsupervised domain adaptation.