Search

Video quality assessment (VQA) task is an ongoing small sample learning problem due to the costly effort required for manual annotation. Since existing VQA datasets are of limited scale, prior research tries to leverage models pre-trained on ImageNet to mitigate this kind of shortage. Nonetheless, these well-trained models targeting on image classification task can be sub-optimal when applied on VQA data from a significantly different domain. In this paper, we make the first attempt to perform self-supervised pre-training for VQA task built upon contrastive learning method, targeting at exploiting the plentiful unlabeled video data to learn feature representation in a simple-yet-effective way. Specifically, we implement this idea by first generating distorted video samples with diverse distortion characteristics and visual contents based on the proposed distortion augmentation strategy. Afterwards, we conduct contrastive learning to capture quality-aware information by maximizing the agreement on feature representations of future frames and their corresponding predictions in the embedding space. In addition, we further introduce distortion prediction task as an additional learning objective to push the model towards discriminating different distortion categories of the input video. Solving these prediction tasks jointly with the contrastive learning not only provides stronger surrogate supervision signals, but also learns the shared knowledge among the prediction tasks. Extensive experiments demonstrate that our approach sets a new state-of-the-art in self-supervised learning for VQA task. Our results also underscore that the learned pre-trained model can significantly benefit the existing learning based VQA models. Source code is available at https://github.com/cpf0079/CSPT.

Referring Expression Comprehension (REC) aims to localize an image region of a given object described by a natural-language expression. While promising performance has been demonstrated, existing REC algorithms make a strong assumption that training data feeding into a model are given upfront, which degrades its practicality for real-world scenarios. In this paper, we propose Continual Referring Expression Comprehension (CREC), a new setting for REC, where a model is learning on a stream of incoming tasks. In order to continuously improve the model on sequential tasks without forgetting prior learned knowledge and without repeatedly re-training from a scratch, we propose an effective baseline method named Dual Modular Memorization (DMM), which alleviates the problem of catastrophic forgetting by two memorization modules: Implicit-Memory and Explicit-Memory. Specifically, the former module aims to constrain drastic changes to important parameters learned on old tasks when learning a new task; while the latter module maintains a buffer pool to dynamically select and store representative samples of each seen task for future rehearsal. We create three benchmarks for the new CREC setting, by respectively re-splitting three widely-used REC datasets RefCOCO, RefCOCO+ and RefCOCOg into sequential tasks. Extensive experiments on the constructed benchmarks demonstrate that our DMM method significantly outperforms other alternatives, based on two popular REC backbones. We make the source code and benchmarks publicly available to foster future progress in this field: https://github.com/zackschen/DMM.

Skeleton-based action recognition is widely used in varied areas, e.g., surveillance and human-machine interaction. Existing models are mainly learned in a supervised manner, thus heavily depending on large-scale labeled data which could be infeasible when labels are prohibitively expensive. In this paper, we propose a novel Contrast-Reconstruction Representation Learning network (CRRL) that simultaneously captures postures and motion dynamics for unsupervised skeleton-based action recognition. It mainly consists of three parts: Sequence Reconstructor, Contrastive Motion Learner, and Information Fuser. The Sequence Reconstructor learns representation from skeleton coordinate sequence via reconstruction, thus the learned representation tends to focus on trivial postural coordinates and be hesitant in motion learning. To enhance the learning of motions, the Contrastive Motion Learner performs contrastive learning between the representations learned from coordinate sequence and additional velocity sequence, respectively. Finally, in the Information Fuser, we explore varied strategies to combine the Sequence Reconstructor and Contrastive Motion Learner, and propose to capture postures and motions simultaneously via a knowledge-distillation based fusion strategy that transfers the motion learning from the Contrastive Motion Learner to the Sequence Reconstructor. Experimental results on several benchmarks, i.e., NTU RGB+D 60, NTU RGB+D 120, CMU mocap, and NW-UCLA, demonstrate the promise of the proposed CRRL method by far outperforming state-of-the-art approaches., Publised in IEEE TIP. (https://ieeexplore.ieee.org/document/9901454)

With the help of convolutional neural networks (CNNs), deep learning-based methods have achieved remarkable performance in face super-resolution (FSR) task. Despite their success, most of the existing methods neglect non-local correlations of face images, leaving much room for improvement. In this paper, we introduce a novel end-to-end trainable attention-driven graph neural network (AD-GNN) for more discriminative feature extraction and feature relation modeling. This is achieved by two major components. The first component is a cross-scale dynamic graph (CDG) block. The CDG block considers cross-scale relationships of patches in distant areas and employs two dynamic graphs to construct enhanced features. The second component is a series of channel attention and spatial dynamic graph (CASDG) blocks. A CASDG block has a channel-wise attention unit and a spatial-aware dynamic graph (SDG) unit. The SDG unit extracts informative features by exploring spatial non-local self-similarity information of the patches using dynamic graph convolution. Using these two components, facial details can be effectively reconstructed with the help of information supplemented by similar but spatially remote patches and structural information of faces. Extensive experiments on two public benchmarks demonstrate the superiority of AD-GNN over the state-of-the-art FSR methods.

Recent deep learning based salient object detection methods with attention mechanisms have made great success. However, existing attention mechanisms can be generally separated into two categories. One part chooses to calculate weights indiscriminately, which yields computational redundancy. While one part focuses randomly on a small part of the images, such as hard attention, resulting in incorrectness owing to insufficiently targeted selection of a subset of tokens. To alleviate these problems, we design a Curiosity-driven Network (CNet) and a Curiosity-driven Learning Algorithm (CLA) based on fragment attention (FA) mechanism newly defined in this paper. FA imitates the process of cognition perception driven by human curiosity, and divides the degree of curiosity into three levels, i.e. curious, a little curious and not curious. These three levels correspond to five saliency degrees, including salient and non-salient, likewise salient and likewise non-salient, completely uncertain. With more knowledge gained by the network, CLA transforms the curiosity degree of each pixel to yield enhanced detail-enriched saliency maps. In order to extract more context-aware information of potential salient objects and make a better foundation for CLA, a high-level feature extraction module (HFEM) is further proposed. Based on the much better high-level features extracted by HFEM, FA can classify the curiosity degree for each pixel more reasonably and accurately. Extensive experiments on five popular datasets clearly demonstrate that our method outperforms the state-of-the-art approaches without any pre-processing operations or post-processing operations.

In this paper, we investigate the problem of abductive visual reasoning (AVR), which requires vision systems to infer the most plausible explanation for visual observations. Unlike previous work which performs visual reasoning on static images or synthesized scenes, we exploit long-term reasoning from instructional videos that contain a wealth of detailed information about the physical world. We conceptualize two tasks for this emerging and challenging topic. The primary task is AVR, which is based on the initial configuration and desired goal from an instructional video, and the model is expected to figure out what is the most plausible sequence of steps to achieve the goal. In order to avoid trivial solutions based on appearance information rather than reasoning, the second task called AVR++ is constructed, which requires the model to answer why the unselected options are less plausible. We introduce a new dataset called VideoABC, which consists of 46,354 unique steps derived from 11,827 instructional videos, formulated as 13,526 abductive reasoning questions with an average reasoning duration of 51 seconds. Through an adversarial hard hypothesis mining algorithm, non-trivial and high-quality problems are generated efficiently and effectively. To achieve human-level reasoning, we propose a Hierarchical Dual Reasoning Network (HDRNet) to capture the long-term dependencies among steps and observations. We establish a benchmark for abductive visual reasoning, and our method set state-of-the-arts on AVR ( ∼ 74 %) and AVR++ ( ∼ 45 %), and humans can easily achieve over 90% accuracy on these two tasks. The large performance gap reveals the limitation of current video understanding models on temporal reasoning and leaves substantial room for future research on this challenging problem. Our dataset and code are available at https://github.com/wl-zhao/VideoABC.

Separating the dominant person from the complex background is significant to the human-related research and photo-editing based applications. Existing segmentation algorithms are either too general to separate the person region accurately, or not capable of achieving real-time speed. In this paper, we introduce the multi-domain learning framework into a novel baseline model to construct the Multi-domain TriSeNet Networks for the real-time single person image segmentation. We first divide training data into different subdomains based on the characteristics of single person images, then apply a multi-branch Feature Fusion Module (FFM) to decouple the networks into the domain-independent and the domain-specific layers. To further enhance the accuracy, a self-supervised learning strategy is proposed to dig out domain relations during training. It helps transfer domain-specific knowledge by improving predictive consistency among different FFM branches. Moreover, we create a large-scale single person image segmentation dataset named MSSP20k, which consists of 22,100 pixel-level annotated images in the real world. The MSSP20k dataset is more complex and challenging than existing public ones in terms of scalability and variety. Experiments show that our Multi-domain TriSeNet outperforms state-of-the-art approaches on both public and the newly built datasets with real-time speed.

Unsupervised domain adaptation (UDA) for person re-identification is challenging because of the huge gap between the source and target domain. A typical self-training method is to use pseudo-labels generated by clustering algorithms to iteratively optimize the model on the target domain. However, a drawback to this is that noisy pseudo-labels generally cause trouble in learning. To address this problem, a mutual learning method by dual networks has been developed to produce reliable soft labels. However, as the two neural networks gradually converge, their complementarity is weakened and they likely become biased towards the same kind of noise. This paper proposes a novel light-weight module, the Attentive WaveBlock (AWB), which can be integrated into the dual networks of mutual learning to enhance the complementarity and further depress noise in the pseudo-labels. Specifically, we first introduce a parameter-free module, the WaveBlock, which creates a difference between features learned by two networks by waving blocks of feature maps differently. Then, an attention mechanism is leveraged to enlarge the difference created and discover more complementary features. Furthermore, two kinds of combination strategies, i.e. pre-attention and post-attention, are explored. Experiments demonstrate that the proposed method achieves state-of-the-art performance with significant improvements on multiple UDA person re-identification tasks. We also prove the generality of the proposed method by applying it to vehicle re-identification and image classification tasks. Our codes and models are available at https://github.com/WangWenhao0716/Attentive-WaveBlock., Comment: Accept to Transaction on Image Processing

Text-based person search aims at retrieving the target person in an image gallery using a descriptive sentence of that person. The core of this task is to calculate a similarity score between the pedestrian image and description, which requires inferring the complex latent correspondence between image sub-regions and textual phrases at different scales. Transformer is an intuitive way to model the complex alignment by its self-attention mechanism. Most previous Transformer-based methods simply concatenate image region features and text features as input and learn a cross-modal representation in a brute force manner. Such weakly supervised learning approaches fail to explicitly build alignment between image region features and text features, causing an inferior feature distribution. In this paper, we present CFLT, Conditional Feature Learning based Transformer. It maps the sub-regions and phrases into a unified latent space and explicitly aligns them by constructing conditional embeddings where the feature of data from one modality is dynamically adjusted based on the data from the other modality. The output of our CFLT is a set of similarity scores for each sub-region or phrase rather than a cross-modal representation. Furthermore, we propose a simple and effective multi-modal re-ranking method named Re-ranking scheme by Visual Conditional Feature (RVCF). Benefit from the visual conditional feature and better feature distribution in our CFLT, the proposed RVCF achieves significant performance improvement. Experimental results show that our CFLT outperforms the state-of-the-art methods by 7.03% in terms of top-1 accuracy and 5.01% in terms of top-5 accuracy on the text-based person search dataset.

Along with the outstanding performance of the deep neural networks (DNNs), considerable research efforts have been devoted to finding ways to understand the decision of DNNs structures. In the computer vision domain, visualizing the attribution map is one of the most intuitive and understandable ways to achieve human-level interpretation. Among them, perturbation-based visualization can explain the "black box" property of the given network by optimizing perturbation masks that alter the network prediction of the target class the most. However, existing perturbation methods could make unexpected changes to network predictions after applying a perturbation mask to the input image, resulting in a loss of robustness and fidelity of the perturbation mechanisms. In this paper, we define class distortion as the unexpected changes of the network prediction during the perturbation process. To handle that, we propose a novel visual interpretation framework, Robust Perturbation, which shows robustness against the unexpected class distortion during the mask optimization. With a new cross-checking mask optimization strategy, our proposed framework perturbs the target prediction of the network while upholding the non-target predictions, providing more reliable and accurate visual explanations. We evaluate our framework on three different public datasets through extensive experiments. Furthermore, we propose a new metric for class distortion evaluation. In both quantitative and qualitative experiments, tackling the class distortion problem turns out to enhance the quality and fidelity of the visual explanation in comparison with the existing perturbation-based methods.