Xiao Pan

Xiao Pan, Peike Li, Zongxin Yang et al.

In this paper, we focus on unsupervised learning for Video Object Segmentation (VOS) which learns visual correspondence (i.e., the similarity between pixel-level features) from unlabeled videos. Previous methods are mainly based on the contrastive learning paradigm, which optimize either in image level or pixel level. Image-level optimization (e.g., the spatially pooled feature of ResNet) learns robust high-level semantics but is sub-optimal since the pixel-level features are optimized implicitly. By contrast, pixel-level optimization is more explicit, however, it is sensitive to the visual quality of training data and is not robust to object deformation. To complementarily perform these two levels of optimization in a unified framework, we propose the In-aNd-Out (INO) generative learning from a purely generative perspective with the help of naturally designed class tokens and patch tokens in Vision Transformer (ViT). Specifically, for image-level optimization, we force the out-view imagination from local to global views on class tokens, which helps capture high-level semantics, and we name it as out-generative learning. As to pixel-level optimization, we perform in-view masked image modeling on patch tokens, which recovers the corrupted parts of an image via inferring its fine-grained structure, and we term it as in-generative learning. To discover the temporal information better, we additionally force the inter-frame consistency from both feature and affinity matrix levels. Extensive experiments on DAVIS-2017 val and YouTube-VOS 2018 val show that our INO outperforms previous state-of-the-art methods by significant margins. Code is available: https://github.com/pansanity666/INO_VOS

Xiao Pan, Zongxin Yang, Jianxin Ma et al.

In this paper, we focus on the task of generalizable neural human rendering which trains conditional Neural Radiance Fields (NeRF) from multi-view videos of different characters. To handle the dynamic human motion, previous methods have primarily used a SparseConvNet (SPC)-based human representation to process the painted SMPL. However, such SPC-based representation i) optimizes under the volatile observation space which leads to the pose-misalignment between training and inference stages, and ii) lacks the global relationships among human parts that is critical for handling the incomplete painted SMPL. Tackling these issues, we present a brand-new framework named TransHuman, which learns the painted SMPL under the canonical space and captures the global relationships between human parts with transformers. Specifically, TransHuman is mainly composed of Transformer-based Human Encoding (TransHE), Deformable Partial Radiance Fields (DPaRF), and Fine-grained Detail Integration (FDI). TransHE first processes the painted SMPL under the canonical space via transformers for capturing the global relationships between human parts. Then, DPaRF binds each output token with a deformable radiance field for encoding the query point under the observation space. Finally, the FDI is employed to further integrate fine-grained information from reference images. Extensive experiments on ZJU-MoCap and H36M show that our TransHuman achieves a significantly new state-of-the-art performance with high efficiency. Project page: https://pansanity666.github.io/TransHuman/

Xiao Pan, Hao Luo, Weihua Chen et al.

We study the backward compatible problem for person re-identification (Re-ID), which aims to constrain the features of an updated new model to be comparable with the existing features from the old model in galleries. Most of the existing works adopt distillation-based methods, which focus on pushing new features to imitate the distribution of the old ones. However, the distillation-based methods are intrinsically sub-optimal since it forces the new feature space to imitate the inferior old feature space. To address this issue, we propose the Ranking-based Backward Compatible Learning (RBCL), which directly optimizes the ranking metric between new features and old features. Different from previous methods, RBCL only pushes the new features to find best-ranking positions in the old feature space instead of strictly alignment, and is in line with the ultimate goal of backward retrieval. However, the sharp sigmoid function used to make the ranking metric differentiable also incurs the gradient vanish issue, therefore stems the ranking refinement during the later period of training. To address this issue, we propose the Dynamic Gradient Reactivation (DGR), which can reactivate the suppressed gradients by adding dynamic computed constant during forward step. To further help targeting the best-ranking positions, we include the Neighbor Context Agents (NCAs) to approximate the entire old feature space during training. Unlike previous works which only test on the in-domain settings, we make the first attempt to introduce the cross-domain settings (including both supervised and unsupervised), which are more meaningful and difficult. The experimental results on all five settings show that the proposed RBCL outperforms previous state-of-the-art methods by large margins under all settings.

Yifei Xiao, T. Y. Yang, Xiao Pan et al.

After an earthquake, it is particularly important to provide the necessary resources on site because a large number of infrastructures need to be repaired or newly constructed. Due to the complex construction environment after the disaster, there are potential safety hazards for human labors working in this environment. With the advancement of robotic technology and artificial intelligent (AI) algorithms, smart robotic technology is the potential solution to provide construction resources after an earthquake. In this paper, the robotic crane with advanced AI algorithms is proposed to provide resources for infrastructure reconstruction after an earthquake. The proximal policy optimization (PPO), a reinforcement learning (RL) algorithm, is implemented for 3D lift path planning when transporting the construction materials. The state and reward function are designed in detail for RL model training. Two models are trained through a loading task in different environments by using PPO algorithm, one considering the influence of obstacles and the other not considering obstacles. Then, the two trained models are compared and evaluated through an unloading task and a loading task in simulation environments. For each task, two different cases are considered. One is that there is no obstacle between the initial position where the construction material is lifted and the target position, and the other is that there are obstacles between the initial position and the target position. The results show that the model that considering the obstacles during training can generate proper actions for the robotic crane to execute so that the crane can automatically transport the construction materials to the desired location with swing suppression, short time consumption and collision avoidance.

Sina Tavasoli, Xiao Pan, T. Y. Yang et al.

Structural columns are the crucial load-carrying components of buildings and bridges. Early detection of column damage is important for the assessment of the residual performance and the prevention of system-level collapse. This research proposes an innovative end-to-end micro aerial vehicles (MAVs)-based approach to automatically scan and inspect columns. First, an MAV-based automatic image collection method is proposed. The MAV is programmed to sense the structural columns and their surrounding environment. During the navigation, the MAV first detects and approaches the structural columns. Then, it starts to collect image data at multiple viewpoints around every detected column. Second, the collected images will be used to assess the damage types and damage locations. Third, the damage state of the structural column will be determined by fusing the evaluation outcomes from multiple camera views. In this study, reinforced concrete (RC) columns are selected to demonstrate the effectiveness of the approach. Experimental results indicate that the proposed MAV-based inspection approach can effectively collect images from multiple viewing angles, and accurately assess critical RC column damages. The approach improves the level of autonomy during the inspection. In addition, the evaluation outcomes are more comprehensive than the existing 2D vision methods. The concept of the proposed inspection approach can be extended to other structural columns such as bridge piers.

Elmira Faraji Zonouz, Xiao Pan, Yu-Cheng Hsu et al.

The detection of masonry damage is essential for preventing potentially disastrous outcomes. Manual inspection can, however, take a long time and be hazardous to human inspectors. Automation of the inspection process using novel computer vision and machine learning algorithms can be a more efficient and safe solution to prevent further deterioration of the masonry structures. Most existing 2D vision-based methods are limited to qualitative damage classification, 2D localization, and in-plane quantification. In this study, we present a 3D vision-based methodology for accurate masonry damage detection, which offers a more robust solution with a greater field of view, depth of vision, and the ability to detect failures in complex environments. First, images of the masonry specimens are collected to generate a 3D point cloud. Second, 3D point clouds processing methods are developed to evaluate the masonry damage. We demonstrate the effectiveness of our approach through experiments on structural masonry components. Our experiments showed the proposed system can effectively classify damage states and localize and quantify critical damage features. The result showed the proposed method can improve the level of autonomy during the inspection of masonry structures.

Xiao Pan, Mingxuan Wang, Liwei Wu et al.

Existing multilingual machine translation approaches mainly focus on English-centric directions, while the non-English directions still lag behind. In this work, we aim to build a many-to-many translation system with an emphasis on the quality of non-English language directions. Our intuition is based on the hypothesis that a universal cross-language representation leads to better multilingual translation performance. To this end, we propose mRASP2, a training method to obtain a single unified multilingual translation model. mRASP2 is empowered by two techniques: a) a contrastive learning scheme to close the gap among representations of different languages, and b) data augmentation on both multiple parallel and monolingual data to further align token representations. For English-centric directions, mRASP2 outperforms existing best unified model and achieves competitive or even better performance than the pre-trained and fine-tuned model mBART on tens of WMT's translation directions. For non-English directions, mRASP2 achieves an improvement of average 10+ BLEU compared with the multilingual Transformer baseline. Code, data and trained models are available at https://github.com/PANXiao1994/mRASP2.

Zehui Lin, Xiao Pan, Mingxuan Wang et al.

We investigate the following question for machine translation (MT): can we develop a single universal MT model to serve as the common seed and obtain derivative and improved models on arbitrary language pairs? We propose mRASP, an approach to pre-train a universal multilingual neural machine translation model. Our key idea in mRASP is its novel technique of random aligned substitution, which brings words and phrases with similar meanings across multiple languages closer in the representation space. We pre-train a mRASP model on 32 language pairs jointly with only public datasets. The model is then fine-tuned on downstream language pairs to obtain specialized MT models. We carry out extensive experiments on 42 translation directions across a diverse settings, including low, medium, rich resource, and as well as transferring to exotic language pairs. Experimental results demonstrate that mRASP achieves significant performance improvement compared to directly training on those target pairs. It is the first time to verify that multiple low-resource language pairs can be utilized to improve rich resource MT. Surprisingly, mRASP is even able to improve the translation quality on exotic languages that never occur in the pre-training corpus. Code, data, and pre-trained models are available at https://github.com/linzehui/mRASP.

Yingjing Xu, Jie Kong, Jiazhi Wang et al.

In this paper, we focus on the task of instruction-based image editing. Previous works like InstructPix2Pix, InstructDiffusion, and SmartEdit have explored end-to-end editing. However, two limitations still remain: First, existing datasets suffer from low resolution, poor background consistency, and overly simplistic instructions. Second, current approaches mainly condition on the text while the rich image information is underexplored, therefore inferior in complex instruction following and maintaining background consistency. Targeting these issues, we first curated the AdvancedEdit dataset using a novel data construction pipeline, formulating a large-scale dataset with high visual quality, complex instructions, and good background consistency. Then, to further inject the rich image information, we introduce a two-stream bridging mechanism utilizing both the textual and visual features reasoned by the powerful Multimodal Large Language Models (MLLM) to guide the image editing process more precisely. Extensive results demonstrate that our approach, InsightEdit, achieves state-of-the-art performance, excelling in complex instruction following and maintaining high background consistency with the original image.

Xiao Pan, Zongxin Yang, Shuai Bai et al.

In this paper, we focus on the One-shot Novel View Synthesis (O-NVS) task which targets synthesizing photo-realistic novel views given only one reference image per scene. Previous One-shot Generalizable Neural Radiance Fields (OG-NeRF) methods solve this task in an inference-time finetuning-free manner, yet suffer the blurry issue due to the encoder-only architecture that highly relies on the limited reference image. On the other hand, recent diffusion-based image-to-3d methods show vivid plausible results via distilling pre-trained 2D diffusion models into a 3D representation, yet require tedious per-scene optimization. Targeting these issues, we propose the GD$^2$-NeRF, a Generative Detail compensation framework via GAN and Diffusion that is both inference-time finetuning-free and with vivid plausible details. In detail, following a coarse-to-fine strategy, GD$^2$-NeRF is mainly composed of a One-stage Parallel Pipeline (OPP) and a 3D-consistent Detail Enhancer (Diff3DE). At the coarse stage, OPP first efficiently inserts the GAN model into the existing OG-NeRF pipeline for primarily relieving the blurry issue with in-distribution priors captured from the training dataset, achieving a good balance between sharpness (LPIPS, FID) and fidelity (PSNR, SSIM). Then, at the fine stage, Diff3DE further leverages the pre-trained image diffusion models to complement rich out-distribution details while maintaining decent 3D consistency. Extensive experiments on both the synthetic and real-world datasets show that GD$^2$-NeRF noticeably improves the details while without per-scene finetuning.

Xiao Pan, Sina Tavasoli, T. Y. Yang et al.

Rapid response to natural disasters such as earthquakes is a crucial element in ensuring the safety of civil infrastructures and minimizing casualties. Traditional manual inspection is labour-intensive, time-consuming, and can be dangerous for inspectors and rescue workers. This paper proposed an autonomous inspection approach for structural damage inspection and survivor detection in the post-disaster building indoor scenario, which incorporates an autonomous navigation method, deep learning-based damage and survivor detection method, and a customized low-cost micro aerial vehicle (MAV) with onboard sensors. Experimental studies in a pseudo-post-disaster office building have shown the proposed methodology can achieve high accuracy in structural damage inspection and survivor detection. Overall, the proposed inspection approach shows great potential to improve the efficiency of existing manual post-disaster building inspection.

Zhongwei Chen, T. Y. Yang, Yifei Xiao et al.

The structural response under the earthquake excitations can be simulated by scaled-down model shake table tests or full-scale model shake table tests. In this paper, adaptive control theory is used as a nonlinear shake table control algorithm which considers the inherent nonlinearity of the shake table system and the Control-Structural Interaction (CSI) effect that the linear controller cannot consider, such as the Proportional-Integral-Derivative (PID) controller. The mass of the specimen can be assumed as an unknown variation and the unknown parameter will be replaced by an estimated value in the proposed control framework. The signal generated by the control law of the adaptive control method will be implemented by a loop-shaping controller. To verify the stability and feasibility of the proposed control framework, a simulation of a bare shake table and experiments with a bare shake table with a two-story frame were carried out. This study randomly selects Earthquake recordings from the Pacific Earthquake Engineering Research Center (PEER) database. The simulation and experimental results show that the proposed control framework can be effectively used in shake table control.

Jinglun Cai, Monica Sunkara, Xilai Li et al.

Masked Language Models (MLMs) have proven to be effective for second-pass rescoring in Automatic Speech Recognition (ASR) systems. In this work, we propose Masked Audio Text Encoder (MATE), a multi-modal masked language model rescorer which incorporates acoustic representations into the input space of MLM. We adopt contrastive learning for effectively aligning the modalities by learning shared representations. We show that using a multi-modal rescorer is beneficial for domain generalization of the ASR system when target domain data is unavailable. MATE reduces word error rate (WER) by 4%-16% on in-domain, and 3%-7% on out-of-domain datasets, over the text-only baseline. Additionally, with very limited amount of training data (0.8 hours), MATE achieves a WER reduction of 8%-23% over the first-pass baseline.

Xiao Pan, Zhizhao Wen, T. Y. Yang

Dynamic analysis of structures subjected to earthquake excitation is a time-consuming process, particularly in the case of extremely small time step required, or in the presence of high geometric and material nonlinearity. Performing parametric studies in such cases is even more tedious. The advancement of computer graphics hardware in recent years enables efficient training of artificial neural networks that are well-known to be capable of learning highly nonlinear mappings. In this study, artificial neural networks are developed with adaptive training algorithms, which enables automatic nodes generation and layers addition. The hyperbolic tangent function is selected as the activation function. Stochastic Gradient Descent and Back Propagation algorithms are adopted to train the networks. The neural networks initiate with a small number of hidden layers and nodes. During training, the performance of the network is continuously tracked, and new nodes or layers are added to the hidden layers if the neural network reaches its capacity. At the end of the training process, the network with appropriate architecture is automatically formed. The performance of the networks has been validated for inelastic shear frames, as well as rocking structures, of which both are first built in finite element program for dynamic analysis to generate training data. Results have shown the developed networks can successfully predict the time-history response of the shear frame and the rock structure subjected to real ground motion records. The efficiency of the proposed neural networks is also examined, which shows the computational time can be reduced by 43% by the neural networks method than FE models. This indicates the trained networks can be utilized to generate rocking spectrums of structures more efficiently which demands a large number of time-history analyses.

Xiao Pan, T. Y. Yang

Structural bolts are critical components used in different structural elements, such as beam-column connections and friction damping devices. The clamping force in structural bolts is highly influenced by the bolt rotation. Much of the existing vision-based research about bolt rotation estimation relies on traditional computer vision algorithms such as Hough Transform to assess static images of bolts. This requires careful image preprocessing, and it may not perform well in the situation of complicated bolt assemblies, or in the presence of surrounding objects and background noise, thus hindering their real-world applications. In this study, an integrated real-time detect-track method, namely RTDT-Bolt, is proposed to monitor the bolt rotation angle. First, a real-time convolutional-neural-networks-based object detector, named YOLOv3-tiny, is established and trained to localize structural bolts. Then, the target-free object tracking algorithm based on optical flow is implemented, to continuously monitor and quantify the rotation of structural bolts. In order to enhance the tracking performance against background noise and potential illumination changes during tracking, the YOLOv3-tiny is integrated with the optical flow tracking algorithm to re-detect the bolts when the tracking gets lost. Extensive parameter studies were conducted to identify optimal tracking performance and examine the potential limitations. The results indicate the RTDT-Bolt method can greatly enhance the tracking performance of bolt rotation, which can achieve over 90% accuracy using the recommended range for the parameters.

Xiao Pan, T. Y. Yang

Reinforced concrete buildings are commonly used around the world. With recent earthquakes worldwide, rapid structural damage inspection and repair cost evaluation are crucial for building owners and policy makers to make informed risk management decisions. To improve the efficiency of such inspection, advanced computer vision techniques based on convolution neural networks have been adopted in recent research to rapidly quantify the damage state of structures. In this paper, an advanced object detection neural network, named YOLO-v2, is implemented which achieves 98.2% and 84.5% average precision in training and testing, respectively. The proposed YOLO-v2 is used in combination with the classification neural network, which improves the identification accuracy for critical damage state of reinforced concrete structures by 7.5%. The improved classification procedures allow engineers to rapidly and more accurately quantify the damage states of the structure, and also localize the critical damage features. The identified damage state can then be integrated with the state-of-the-art performance evaluation framework to quantify the financial losses of critical reinforced concrete buildings. The results can be used by the building owners and decision makers to make informed risk management decisions immediately after the strong earthquake shaking. Hence, resources can be allocated rapidly to improve the resiliency of the community.

Liwei Wu, Xiao Pan, Zehui Lin et al.

This paper describes our VolcTrans system on WMT20 shared news translation task. We participated in 8 translation directions. Our basic systems are based on Transformer, with several variants (wider or deeper Transformers, dynamic convolutions). The final system includes text pre-process, data selection, synthetic data generation, advanced model ensemble, and multilingual pre-training.

Jiawei Zhang, Xiao Pan, Moyin Li et al.

Bicycle-sharing systems, which can provide shared bike usage services for the public, have been launched in many big cities. In bicycle-sharing systems, people can borrow and return bikes at any stations in the service region very conveniently. Therefore, bicycle-sharing systems are normally used as a short-distance trip supplement for private vehicles as well as regular public transportation. Meanwhile, for stations located at different places in the service region, the bike usages can be quite skewed and imbalanced. Some stations have too many incoming bikes and get jammed without enough docks for upcoming bikes, while some other stations get empty quickly and lack enough bikes for people to check out. Therefore, inferring the potential destinations and arriving time of each individual trip beforehand can effectively help the service providers schedule manual bike re-dispatch in advance. In this paper, we will study the individual trip prediction problem for bicycle-sharing systems. To address the problem, we study a real-world bicycle-sharing system and analyze individuals' bike usage behaviors first. Based on the analysis results, a new trip destination prediction and trip duration inference model will be introduced. Experiments conducted on a real-world bicycle-sharing system demonstrate the effectiveness of the proposed model.