Hao Chen

Ning Han, Xun Yang, Ee-Peng Lim et al.

Cross-modal video retrieval aims to retrieve the semantically relevant videos given a text as a query, and is one of the fundamental tasks in Multimedia. Most of top-performing methods primarily leverage Visual Transformer (ViT) to extract video features [1, 2, 3], suffering from high computational complexity of ViT especially for encoding long videos. A common and simple solution is to uniformly sample a small number (say, 4 or 8) of frames from the video (instead of using the whole video) as input to ViT. The number of frames has a strong influence on the performance of ViT, e.g., using 8 frames performs better than using 4 frames yet needs more computational resources, resulting in a trade-off. To get free from this trade-off, this paper introduces an automatic video compression method based on a bilevel optimization program (BOP) consisting of both model-level (i.e., base-level) and frame-level (i.e., meta-level) optimizations. The model-level learns a cross-modal video retrieval model whose input is the "compressed frames" learned by frame-level optimization. In turn, the frame-level optimization is through gradient descent using the meta loss of video retrieval model computed on the whole video. We call this BOP method as well as the "compressed frames" as Meta-Optimized Frames (MOF). By incorporating MOF, the video retrieval model is able to utilize the information of whole videos (for training) while taking only a small number of input frames in actual implementation. The convergence of MOF is guaranteed by meta gradient descent algorithms. For evaluation, we conduct extensive experiments of cross-modal video retrieval on three large-scale benchmarks: MSR-VTT, MSVD, and DiDeMo. Our results show that MOF is a generic and efficient method to boost multiple baseline methods, and can achieve a new state-of-the-art performance.

Hao Chen, Feihong Shen

Most of existing RGB-D salient object detection (SOD) methods follow the CNN-based paradigm, which is unable to model long-range dependencies across space and modalities due to the natural locality of CNNs. Here we propose the Hierarchical Cross-modal Transformer (HCT), a new multi-modal transformer, to tackle this problem. Unlike previous multi-modal transformers that directly connecting all patches from two modalities, we explore the cross-modal complementarity hierarchically to respect the modality gap and spatial discrepancy in unaligned regions. Specifically, we propose to use intra-modal self-attention to explore complementary global contexts, and measure spatial-aligned inter-modal attention locally to capture cross-modal correlations. In addition, we present a Feature Pyramid module for Transformer (FPT) to boost informative cross-scale integration as well as a consistency-complementarity module to disentangle the multi-modal integration path and improve the fusion adaptivity. Comprehensive experiments on a large variety of public datasets verify the efficacy of our designs and the consistent improvement over state-of-the-art models.

Haotai Liang, Mengran Shi, Chen Dong et al.

The key feature of model-driven semantic communication is the propagation of the model. The semantic model component (SMC) is designed to drive the intelligent model to transmit in the physical channel, allowing the intelligence to flow through the networks. According to the characteristics of neural networks with common and individual model parameters, this paper designs the cross-source-domain and cross-task semantic component model. Considering that the basic model is deployed on the edge node, the large server node updates the edge node by transmitting only the semantic component model to the edge node so that the edge node can handle different sources and different tasks. In addition, this paper also discusses how channel noise affects the performance of the model and proposes methods of injection noise and regularization to improve the noise resistance of the model. Experiments show that SMCs use smaller model parameters to achieve cross-source, cross-task functionality while maintaining performance and improving the model's tolerance to noise. Finally, a component transfer-based unmanned vehicle tracking prototype was implemented to verify the feasibility of model components in practical applications.

Hao Chen, Yuqi Hou, Chenyuan Qu et al.

Human perception of the world is shaped by a multitude of viewpoints and modalities. While many existing datasets focus on scene understanding from a certain perspective (e.g. egocentric or third-person views), our dataset offers a panoptic perspective (i.e. multiple viewpoints with multiple data modalities). Specifically, we encapsulate third-person panoramic and front views, as well as egocentric monocular/binocular views with rich modalities including video, multi-channel audio, directional binaural delay, location data and textual scene descriptions within each scene captured, presenting comprehensive observation of the world. Figure 1 offers a glimpse of all 28 scene categories of our 360+x dataset. To the best of our knowledge, this is the first database that covers multiple viewpoints with multiple data modalities to mimic how daily information is accessed in the real world. Through our benchmark analysis, we presented 5 different scene understanding tasks on the proposed 360+x dataset to evaluate the impact and benefit of each data modality and perspective in panoptic scene understanding. We hope this unique dataset could broaden the scope of comprehensive scene understanding and encourage the community to approach these problems from more diverse perspectives.

Wen Wang, Qiuyu Wang, Kecheng Zheng et al.

We propose Framer for interactive frame interpolation, which targets producing smoothly transitioning frames between two images as per user creativity. Concretely, besides taking the start and end frames as inputs, our approach supports customizing the transition process by tailoring the trajectory of some selected keypoints. Such a design enjoys two clear benefits. First, incorporating human interaction mitigates the issue arising from numerous possibilities of transforming one image to another, and in turn enables finer control of local motions. Second, as the most basic form of interaction, keypoints help establish the correspondence across frames, enhancing the model to handle challenging cases (e.g., objects on the start and end frames are of different shapes and styles). It is noteworthy that our system also offers an "autopilot" mode, where we introduce a module to estimate the keypoints and refine the trajectory automatically, to simplify the usage in practice. Extensive experimental results demonstrate the appealing performance of Framer on various applications, such as image morphing, time-lapse video generation, cartoon interpolation, etc. The code, the model, and the interface will be released to facilitate further research.

Hao Chen, Jiaze Wang, Ziyu Guo et al.

Sign language recognition (SLR) plays a vital role in facilitating communication for the hearing-impaired community. SLR is a weakly supervised task where entire videos are annotated with glosses, making it challenging to identify the corresponding gloss within a video segment. Recent studies indicate that the main bottleneck in SLR is the insufficient training caused by the limited availability of large-scale datasets. To address this challenge, we present SignVTCL, a multi-modal continuous sign language recognition framework enhanced by visual-textual contrastive learning, which leverages the full potential of multi-modal data and the generalization ability of language model. SignVTCL integrates multi-modal data (video, keypoints, and optical flow) simultaneously to train a unified visual backbone, thereby yielding more robust visual representations. Furthermore, SignVTCL contains a visual-textual alignment approach incorporating gloss-level and sentence-level alignment to ensure precise correspondence between visual features and glosses at the level of individual glosses and sentence. Experimental results conducted on three datasets, Phoenix-2014, Phoenix-2014T, and CSL-Daily, demonstrate that SignVTCL achieves state-of-the-art results compared with previous methods.

Diego Monteiro, Hao Chen, Hai-Ning Liang et al.

In virtual reality (VR) games, playability and immersion levels are important because they affect gameplay, enjoyment, and performance. However, they can be adversely affected by VR sickness (VRS) symptoms. VRS can be minimized by manipulating users' perception of the virtual environment via the head-mounted display (HMD). One extreme example is the Teleport mitigation technique, which lets users navigate discretely, skipping sections of the virtual space. Other techniques are less extreme but still rely on controlling what and how much users see via the HMD. This research examines the effect on players' performance and gameplay of these mitigation techniques in fast-paced VR games. Our focus is on two types of visual reduction techniques. This study aims to identify specifically the trade-offs these techniques have in a first-person shooter game regarding immersion, performance, and VRS. The main contributions in this paper are (1) a deeper understanding of one of the most popular techniques (Teleport) when it comes to gameplay; (2) the replication and validation of a novel VRS mitigation technique based on visual reduction; and (3) a comparison of their effect on players' performance and gameplay.

Yongjian Deng, Hao Chen, Hai Liu et al.

Event cameras attract researchers' attention due to their low power consumption, high dynamic range, and extremely high temporal resolution. Learning models on event-based object classification have recently achieved massive success by accumulating sparse events into dense frames to apply traditional 2D learning methods. Yet, these approaches necessitate heavy-weight models and are with high computational complexity due to the redundant information introduced by the sparse-to-dense conversion, limiting the potential of event cameras on real-life applications. This study aims to address the core problem of balancing accuracy and model complexity for event-based classification models. To this end, we introduce a novel graph representation for event data to exploit their sparsity better and customize a lightweight voxel graph convolutional neural network (\textit{EV-VGCNN}) for event-based classification. Specifically, (1) using voxel-wise vertices rather than previous point-wise inputs to explicitly exploit regional 2D semantics of event streams while keeping the sparsity;(2) proposing a multi-scale feature relational layer (\textit{MFRL}) to extract spatial and motion cues from each vertex discriminatively concerning its distances to neighbors. Comprehensive experiments show that our model can advance state-of-the-art classification accuracy with extremely low model complexity (merely 0.84M parameters).

Jialin Wang, Hai-Ning Liang, Diego Monteiro et al.

Virtual Reality (VR) technology has been proliferating in the last decade, especially in the last few years. However, Simulator Sickness (SS) still represents a significant problem for its wider adoption. Currently, the most common way to detect SS is using the Simulator Sickness Questionnaire (SSQ). SSQ is a subjective measurement and is inadequate for real-time applications such as VR games. This research aims to investigate how to use machine learning techniques to detect SS based on in-game characters' and users' physiological data during gameplay in VR games. To achieve this, we designed an experiment to collect such data with three types of games. We trained a Long Short-Term Memory neural network with the dataset eye-tracking and character movement data to detect SS in real-time. Our results indicate that, in VR games, our model is an accurate and efficient way to detect SS in real-time.

Diego Monteiro, Hai-Ning Liang, Jialin Wang et al.

The amount of interest in Virtual Reality (VR) research has significantly increased over the past few years, both in academia and industry. The release of commercial VR Head-Mounted Displays (HMDs) has been a major contributing factor. However, there is still much to be learned, especially how views and input techniques, as well as their interaction, affect the VR experience. There is little work done on First-Person Shooter (FPS) games in VR, and those few studies have focused on a single aspect of VR FPS. They either focused on the view, e.g., comparing VR to a typical 2D display or on the controller types. To the best of our knowledge, there are no studies investigating variations of 2D/3D views in HMDs, controller types, and their interactions. As such, it is challenging to distinguish findings related to the controller type from those related to the view. If a study does not control for the input method and finds that 2D displays lead to higher performance than VR, we cannot generalize the results because of the confounding variables. To understand their interaction, we propose to analyze in more depth, whether it is the view (2D vs. 3D) or the way it is controlled that gives the platforms their respective advantages. To study the effects of the 2D/3D views, we created a 2D visual technique, PlaneFrame, that was applied inside the VR headset. Our results show that the controller type can have a significant positive impact on performance, immersion, and simulator sickness when associated with a 2D view. They further our understanding of the interactions that controllers and views have and demonstrate that comparisons are highly dependent on how both factors go together. Further, through a series of three experiments, we developed a technique that can lead to a substantial performance, a good level of immersion, and can minimize the level of simulator sickness.

Hao Chen, Youfu Li

The goal of this work is to present a systematic solution for RGB-D salient object detection, which addresses the following three aspects with a unified framework: modal-specific representation learning, complementary cue selection and cross-modal complement fusion. To learn discriminative modal-specific features, we propose a hierarchical cross-modal distillation scheme, in which the well-learned source modality provides supervisory signals to facilitate the learning process for the new modality. To better extract the complementary cues, we formulate a residual function to incorporate complements from the paired modality adaptively. Furthermore, a top-down fusion structure is constructed for sufficient cross-modal interactions and cross-level transmissions. The experimental results demonstrate the effectiveness of the proposed cross-modal distillation scheme in zero-shot saliency detection and pre-training on a new modality, as well as the advantages in selecting and fusing cross-modal/cross-level complements.

Hao Chen, Y. F. Li, Dan Su

In this work, we propose to utilize Convolutional Neural Networks to boost the performance of depth-induced salient object detection by capturing the high-level representative features for depth modality. We formulate the depth-induced saliency detection as a CNN-based cross-modal transfer problem to bridge the gap between the "data-hungry" nature of CNNs and the unavailability of sufficient labeled training data in depth modality. In the proposed approach, we leverage the auxiliary data from the source modality effectively by training the RGB saliency detection network to obtain the task-specific pre-understanding layers for the target modality. Meanwhile, we exploit the depth-specific information by pre-training a modality classification network that encourages modal-specific representations during the optimizing course. Thus, it could make the feature representations of the RGB and depth modalities as discriminative as possible. These two modules are pre-trained independently and then stitched to initialize and optimize the eventual depth-induced saliency detection model. Experiments demonstrate the effectiveness of the proposed novel pre-training strategy as well as the significant and consistent improvements of the proposed approach over other state-of-the-art methods.