Zejian Yuan

Lei Wang, Zejian Yuan, Yao Lu et al.

Scene Graph Generation (SGG) as a critical task in image understanding, facing the challenge of head-biased prediction caused by the long-tail distribution of predicates. However, current unbiased SGG methods can easily prioritize improving the prediction of tail predicates while ignoring the substantial sacrifice in the prediction of head predicates, leading to a shift from head bias to tail bias. To address this issue, we propose a model-agnostic Head-Tail Collaborative Learning (HTCL) network that includes head-prefer and tail-prefer feature representation branches that collaborate to achieve accurate recognition of both head and tail predicates. We also propose a self-supervised learning approach to enhance the prediction ability of the tail-prefer feature representation branch by constraining tail-prefer predicate features. Specifically, self-supervised learning converges head predicate features to their class centers while dispersing tail predicate features as much as possible through contrast learning and head center loss. We demonstrate the effectiveness of our HTCL by applying it to various SGG models on VG150, Open Images V6 and GQA200 datasets. The results show that our method achieves higher mean Recall with a minimal sacrifice in Recall and achieves a new state-of-the-art overall performance. Our code is available at https://github.com/wanglei0618/HTCL.

Zhihao Guan, Ruixin Liu, Zejian Yuan et al.

As one of the basic while vital technologies for HD map construction, 3D lane detection is still an open problem due to varying visual conditions, complex typologies, and strict demands for precision. In this paper, an end-to-end flexible and hierarchical lane detector is proposed to precisely predict 3D lane lines from point clouds. Specifically, we design a hierarchical network predicting flexible representations of lane shapes at different levels, simultaneously collecting global instance semantics and avoiding local errors. In the global scope, we propose to regress parametric curves w.r.t adaptive axes that help to make more robust predictions towards complex scenes, while in the local vision the structure of lane segment is detected in each of the dynamic anchor cells sampled along the global predicted curves. Moreover, corresponding global and local shape matching losses and anchor cell generation strategies are designed. Experiments on two datasets show that we overwhelm current top methods under high precision standards, and full ablation studies also verify each part of our method. Our codes will be released at https://github.com/Doo-do/FHLD.

Xinyi He, Mengyu Zhou, Mingjie Zhou et al. · stanford

Tabular data analysis is performed every day across various domains. It requires an accurate understanding of field semantics to correctly operate on table fields and find common patterns in daily analysis. In this paper, we introduce the AnaMeta dataset, a collection of 467k tables with derived supervision labels for four types of commonly used field metadata: measure/dimension dichotomy, common field roles, semantic field type, and default aggregation function. We evaluate a wide range of models for inferring metadata as the benchmark. We also propose a multi-encoder framework, called KDF, which improves the metadata understanding capability of tabular models by incorporating distribution and knowledge information. Furthermore, we propose four interfaces for incorporating field metadata into downstream analysis tasks.

Ruijin Liu, Ning Lu, Dapeng Chen et al.

We present PBFormer, an efficient yet powerful scene text detector that unifies the transformer with a novel text shape representation Polynomial Band (PB). The representation has four polynomial curves to fit a text's top, bottom, left, and right sides, which can capture a text with a complex shape by varying polynomial coefficients. PB has appealing features compared with conventional representations: 1) It can model different curvatures with a fixed number of parameters, while polygon-points-based methods need to utilize a different number of points. 2) It can distinguish adjacent or overlapping texts as they have apparent different curve coefficients, while segmentation-based or points-based methods suffer from adhesive spatial positions. PBFormer combines the PB with the transformer, which can directly generate smooth text contours sampled from predicted curves without interpolation. A parameter-free cross-scale pixel attention (CPA) module is employed to highlight the feature map of a suitable scale while suppressing the other feature maps. The simple operation can help detect small-scale texts and is compatible with the one-stage DETR framework, where no postprocessing exists for NMS. Furthermore, PBFormer is trained with a shape-contained loss, which not only enforces the piecewise alignment between the ground truth and the predicted curves but also makes curves' positions and shapes consistent with each other. Without bells and whistles about text pre-training, our method is superior to the previous state-of-the-art text detectors on the arbitrary-shaped text datasets.

Ruixin Liu, Zejian Yuan

Monocular 3D lane detection is challenged by aleatoric uncertainty arising from inherent observation noise. Existing methods rely on simplified geometric assumptions, such as independent point predictions or global planar modeling, failing to capture structural variations and aleatoric uncertainty in real-world scenarios. In this paper, we propose MonoUnc, a bird's-eye view (BEV)-free 3D lane detector that explicitly models aleatoric uncertainty informed by local lane structures. Specifically, 3D lanes are projected onto the front-view (FV) space and approximated by parametric curves. Guided by curve predictions, curve-point query embeddings are dynamically generated for lane point predictions in 3D space. Each segment formed by two adjacent points is modeled as a 3D Gaussian, parameterized by the local structure and uncertainty estimations. Accordingly, a novel 3D Gaussian matching loss is designed to constrain these parameters jointly. Experiments on the ONCE-3DLanes and OpenLane datasets demonstrate that MonoUnc outperforms previous state-of-the-art (SoTA) methods across all benchmarks under stricter evaluation criteria. Additionally, we propose two comprehensive evaluation metrics for ONCE-3DLanes, calculating the average and maximum bidirectional Chamfer distances to quantify global and local errors. Codes are released at https://github.com/lrx02/MonoUnc.

Ruijin Liu, Dapeng Chen, Tie Liu et al.

Detecting 3D lanes from the camera is a rising problem for autonomous vehicles. In this task, the correct camera pose is the key to generating accurate lanes, which can transform an image from perspective-view to the top-view. With this transformation, we can get rid of the perspective effects so that 3D lanes would look similar and can accurately be fitted by low-order polynomials. However, mainstream 3D lane detectors rely on perfect camera poses provided by other sensors, which is expensive and encounters multi-sensor calibration issues. To overcome this problem, we propose to predict 3D lanes by estimating camera pose from a single image with a two-stage framework. The first stage aims at the camera pose task from perspective-view images. To improve pose estimation, we introduce an auxiliary 3D lane task and geometry constraints to benefit from multi-task learning, which enhances consistencies between 3D and 2D, as well as compatibility in the above two tasks. The second stage targets the 3D lane task. It uses previously estimated pose to generate top-view images containing distance-invariant lane appearances for predicting accurate 3D lanes. Experiments demonstrate that, without ground truth camera pose, our method outperforms the state-of-the-art perfect-camera-pose-based methods and has the fewest parameters and computations. Codes are available at https://github.com/liuruijin17/CLGo.

Ruijin Liu, Zejian Yuan, Tie Liu et al.

Lane detection, the process of identifying lane markings as approximated curves, is widely used for lane departure warning and adaptive cruise control in autonomous vehicles. The popular pipeline that solves it in two steps -- feature extraction plus post-processing, while useful, is too inefficient and flawed in learning the global context and lanes' long and thin structures. To tackle these issues, we propose an end-to-end method that directly outputs parameters of a lane shape model, using a network built with a transformer to learn richer structures and context. The lane shape model is formulated based on road structures and camera pose, providing physical interpretation for parameters of network output. The transformer models non-local interactions with a self-attention mechanism to capture slender structures and global context. The proposed method is validated on the TuSimple benchmark and shows state-of-the-art accuracy with the most lightweight model size and fastest speed. Additionally, our method shows excellent adaptability to a challenging self-collected lane detection dataset, showing its powerful deployment potential in real applications. Codes are available at https://github.com/liuruijin17/LSTR.

Xinyi He, Mengyu Zhou, Xinrun Xu et al.

Tabular data analysis is crucial in various fields, and large language models show promise in this area. However, current research mostly focuses on rudimentary tasks like Text2SQL and TableQA, neglecting advanced analysis like forecasting and chart generation. To address this gap, we developed the Text2Analysis benchmark, incorporating advanced analysis tasks that go beyond the SQL-compatible operations and require more in-depth analysis. We also develop five innovative and effective annotation methods, harnessing the capabilities of large language models to enhance data quality and quantity. Additionally, we include unclear queries that resemble real-world user questions to test how well models can understand and tackle such challenges. Finally, we collect 2249 query-result pairs with 347 tables. We evaluate five state-of-the-art models using three different metrics and the results show that our benchmark presents introduces considerable challenge in the field of tabular data analysis, paving the way for more advanced research opportunities.

Zhupeng Ye, Yinqi Li, Zejian Yuan

Wide-range and fine-grained vehicle detection plays a critical role in enabling active safety features in intelligent driving systems. However, existing vehicle detection methods based on rectangular bounding boxes (BBox) often struggle with perceiving wide-range objects, especially small objects at long distances. And BBox expression cannot provide detailed geometric shape and pose information of vehicles. This paper proposes a novel wide-range Pseudo-3D Vehicle Detection method based on images from a single camera and incorporates efficient learning methods. This model takes a spliced image as input, which is obtained by combining two sub-window images from a high-resolution image. This image format maximizes the utilization of limited image resolution to retain essential information about wide-range vehicle objects. To detect pseudo-3D objects, our model adopts specifically designed detection heads. These heads simultaneously output extended BBox and Side Projection Line (SPL) representations, which capture vehicle shapes and poses, enabling high-precision detection. To further enhance the performance of detection, a joint constraint loss combining both the object box and SPL is designed during model training, improving the efficiency, stability, and prediction accuracy of the model. Experimental results on our self-built dataset demonstrate that our model achieves favorable performance in wide-range pseudo-3D vehicle detection across multiple evaluation metrics. Our demo video has been placed at https://www.youtube.com/watch?v=1gk1PmsQ5Q8.

Xinyi He, Yihao Liu, Mengyu Zhou et al.

Tabular data are crucial in many fields and their understanding by large language models (LLMs) under high parameter efficiency paradigm is important. However, directly applying parameter-efficient fine-tuning (PEFT) techniques to tabular tasks presents significant challenges, particularly in terms of better table serialization and the representation of two-dimensional structured information within a one-dimensional sequence. To address this, we propose TableLoRA, a module designed to improve LLMs' understanding of table structure during PEFT. It incorporates special tokens for serializing tables with special token encoder and uses 2D LoRA to encode low-rank information on cell positions. Experiments on four tabular-related datasets demonstrate that TableLoRA consistently outperforms vanilla LoRA and surpasses various table encoding methods tested in control experiments. These findings reveal that TableLoRA, as a table-specific LoRA, enhances the ability of LLMs to process tabular data effectively, especially in low-parameter settings, demonstrating its potential as a robust solution for handling table-related tasks.

Bin Dou, Tianyu Zhang, Zhaohui Wang et al.

Zero-shot neural scene segmentation, which reconstructs 3D neural segmentation field without manual annotations, serves as an effective way for scene understanding. However, existing models, especially the efficient 3D Gaussian-based methods, struggle to produce compact segmentation results. This issue stems primarily from their redundant learnable attributes assigned on individual Gaussians, leading to a lack of robustness against the 3D-inconsistencies in zero-shot generated raw labels. To address this problem, our work, named Compact Segmented 3D Gaussians (CoSegGaussians), proposes the Feature Unprojection and Fusion module as the segmentation field, which utilizes a shallow decoder generalizable for all Gaussians based on high-level features. Specifically, leveraging the learned Gaussian geometric parameters, semantic-aware image-based features are introduced into the scene via our unprojection technique. The lifted features, together with spatial information, are fed into the multi-scale aggregation decoder to generate segmentation identities for all Gaussians. Furthermore, we design CoSeg Loss to boost model robustness against 3D-inconsistent noises. Experimental results show that our model surpasses baselines on zero-shot semantic segmentation task, improving by ~10% mIoU over the best baseline. Code and more results will be available at https://David-Dou.github.io/CoSegGaussians.

Xinyi He, Jiaru Zou, Yun Lin et al.

Large Language Models have revolutionized code generation ability by converting natural language descriptions into executable code. However, generating complex code within real-world scenarios remains challenging due to intricate structures, subtle bugs, understanding of advanced data types, and lack of supplementary contents. To address these challenges, we introduce the CoCoST framework, which enhances complex code generation by online searching for more information with planned queries and correctness testing for code refinement. Moreover, CoCoST serializes the complex inputs and outputs to improve comprehension and generates test cases to ensure the adaptability for real-world applications. CoCoST is validated through rigorous experiments on the DS-1000 and ClassEval datasets. Experimental results show that CoCoST substantially improves the quality of complex code generation, highlighting its potential to enhance the practicality of LLMs in generating complex code.

Zidong Guo, Zejian Yuan, Chong Zhang et al.

Gaze is the essential manifestation of human attention. In recent years, a series of work has achieved high accuracy in gaze estimation. However, the inter-personal difference limits the reduction of the subject-independent gaze estimation error. This paper proposes an unsupervised method for domain adaptation gaze estimation to eliminate the impact of inter-personal diversity. In domain adaption, we design an embedding representation with prediction consistency to ensure that the linear relationship between gaze directions in different domains remains consistent on gaze space and embedding space. Specifically, we employ source gaze to form a locally linear representation in the gaze space for each target domain prediction. Then the same linear combinations are applied in the embedding space to generate hypothesis embedding for the target domain sample, remaining prediction consistency. The deviation between the target and source domain is reduced by approximating the predicted and hypothesis embedding for the target domain sample. Guided by the proposed strategy, we design Domain Adaptation Gaze Estimation Network(DAGEN), which learns embedding with prediction consistency and achieves state-of-the-art results on both the MPIIGaze and the EYEDIAP datasets.

Ziyang Song, Zejian Yuan, Chong Zhang et al.

In recognition-based action interaction, robots' responses to human actions are often pre-designed according to recognized categories and thus stiff. In this paper, we specify a new Interactive Action Translation (IAT) task which aims to learn end-to-end action interaction from unlabeled interactive pairs, removing explicit action recognition. To enable learning on small-scale data, we propose a Paired-Embedding (PE) method for effective and reliable data augmentation. Specifically, our method first utilizes paired relationships to cluster individual actions in an embedding space. Then two actions originally paired can be replaced with other actions in their respective neighborhood, assembling into new pairs. An Act2Act network based on conditional GAN follows to learn from augmented data. Besides, IAT-test and IAT-train scores are specifically proposed for evaluating methods on our task. Experimental results on two datasets show impressive effects and broad application prospects of our method.

Ziyang Song, Ziyi Yin, Zejian Yuan et al.

Despite the notable progress made in action recognition tasks, not much work has been done in action recognition specifically for human-robot interaction. In this paper, we deeply explore the characteristics of the action recognition task in interaction scenarios and propose an attention-oriented multi-level network framework to meet the need for real-time interaction. Specifically, a Pre-Attention network is employed to roughly focus on the interactor in the scene at low resolution firstly and then perform fine-grained pose estimation at high resolution. The other compact CNN receives the extracted skeleton sequence as input for action recognition, utilizing attention-like mechanisms to capture local spatial-temporal patterns and global semantic information effectively. To evaluate our approach, we construct a new action dataset specially for the recognition task in interaction scenarios. Experimental results on our dataset and high efficiency (112 fps at 640 x 480 RGBD) on the mobile computing platform (Nvidia Jetson AGX Xavier) demonstrate excellent applicability of our method on action recognition in real-time human-robot interaction.

Badong Chen, Yuqing Xie, Xin Wang et al.

As a novel similarity measure that is defined as the expectation of a kernel function between two random variables, correntropy has been successfully applied in robust machine learning and signal processing to combat large outliers. The kernel function in correntropy is usually a zero-mean Gaussian kernel. In a recent work, the concept of mixture correntropy (MC) was proposed to improve the learning performance, where the kernel function is a mixture Gaussian kernel, namely a linear combination of several zero-mean Gaussian kernels with different widths. In both correntropy and mixture correntropy, the center of the kernel function is, however, always located at zero. In the present work, to further improve the learning performance, we propose the concept of multi-kernel correntropy (MKC), in which each component of the mixture Gaussian kernel can be centered at a different location. The properties of the MKC are investigated and an efficient approach is proposed to determine the free parameters in MKC. Experimental results show that the learning algorithms under the maximum multi-kernel correntropy criterion (MMKCC) can outperform those under the original maximum correntropy criterion (MCC) and the maximum mixture correntropy criterion (MMCC).

Zhihao Zhu, Zhan Xue, Zejian Yuan

Recent progress on deep learning has made it possible to automatically transform the screenshot of Graphic User Interface (GUI) into code by using the encoder-decoder framework. While the commonly adopted image encoder (e.g., CNN network), might be capable of extracting image features to the desired level, interpreting these abstract image features into hundreds of tokens of code puts a particular challenge on the decoding power of the RNN-based code generator. Considering the code used for describing GUI is usually hierarchically structured, we propose a new attention-based hierarchical code generation model, which can describe GUI images in a finer level of details, while also being able to generate hierarchically structured code in consistency with the hierarchical layout of the graphic elements in the GUI. Our model follows the encoder-decoder framework, all the components of which can be trained jointly in an end-to-end manner. The experimental results show that our method outperforms other current state-of-the-art methods on both a publicly available GUI-code dataset as well as a dataset established by our own.

Yuanliu Liu, Zejian Yuan

Intrinsic image decomposition is a severely under-constrained problem. User interactions can help to reduce the ambiguity of the decomposition considerably. The traditional way of user interaction is to draw scribbles that indicate regions with constant reflectance or shading. However the effect scopes of the scribbles are quite limited, so dozens of scribbles are often needed to rectify the whole decomposition, which is time consuming. In this paper we propose an efficient way of user interaction that users need only to annotate the color composition of the image. Color composition reveals the global distribution of reflectance, so it can help to adapt the whole decomposition directly. We build a generative model of the process that the albedo of the material produces both the reflectance through imaging and the color labels by color naming. Our model fuses effectively the physical properties of image formation and the top-down information from human color perception. Experimental results show that color naming can improve the performance of intrinsic image decomposition, especially in cleaning the shadows left in reflectance and solving the color constancy problem.

Yuanliu Liu, Ang Li, Zejian Yuan et al.

We address the problem of decomposing a single image into reflectance and shading. The difficulty comes from the fact that the components of image---the surface albedo, the direct illumination, and the ambient illumination---are coupled heavily in observed image. We propose to infer the shading by ordering pixels by their relative brightness, without knowing the absolute values of the image components beforehand. The pairwise shading orders are estimated in two ways: brightness order and low-order fittings of local shading field. The brightness order is a non-local measure, which can be applied to any pair of pixels including those whose reflectance and shading are both different. The low-order fittings are used for pixel pairs within local regions of smooth shading. Together, they can capture both global order structure and local variations of the shading. We propose a Consistency-aware Selective Fusion (CSF) to integrate the pairwise orders into a globally consistent order. The iterative selection process solves the conflicts between the pairwise orders obtained by different estimation methods. Inconsistent or unreliable pairwise orders will be automatically excluded from the fusion to avoid polluting the global order. Experiments on the MIT Intrinsic Image dataset show that the proposed model is effective at recovering the shading including deep shadows. Our model also works well on natural images from the IIW dataset, the UIUC Shadow dataset and the NYU-Depth dataset, where the colors of direct lights and ambient lights are quite different.

Dapeng Chen, Hongsheng Li, Xihui Liu et al.

Person re-identification is an important task that requires learning discriminative visual features for distinguishing different person identities. Diverse auxiliary information has been utilized to improve the visual feature learning. In this paper, we propose to exploit natural language description as additional training supervisions for effective visual features. Compared with other auxiliary information, language can describe a specific person from more compact and semantic visual aspects, thus is complementary to the pixel-level image data. Our method not only learns better global visual feature with the supervision of the overall description but also enforces semantic consistencies between local visual and linguistic features, which is achieved by building global and local image-language associations. The global image-language association is established according to the identity labels, while the local association is based upon the implicit correspondences between image regions and noun phrases. Extensive experiments demonstrate the effectiveness of employing language as training supervisions with the two association schemes. Our method achieves state-of-the-art performance without utilizing any auxiliary information during testing and shows better performance than other joint embedding methods for the image-language association.

Zhihao Zhu, Zhan Xue, Zejian Yuan

Attention mechanisms have attracted considerable interest in image captioning because of its powerful performance. Existing attention-based models use feedback information from the caption generator as guidance to determine which of the image features should be attended to. A common defect of these attention generation methods is that they lack a higher-level guiding information from the image itself, which sets a limit on selecting the most informative image features. Therefore, in this paper, we propose a novel attention mechanism, called topic-guided attention, which integrates image topics in the attention model as a guiding information to help select the most important image features. Moreover, we extract image features and image topics with separate networks, which can be fine-tuned jointly in an end-to-end manner during training. The experimental results on the benchmark Microsoft COCO dataset show that our method yields state-of-art performance on various quantitative metrics.

Ang Li, Zejian Yuan

Detecting the occlusion from stereo images or video frames is important to many computer vision applications. Previous efforts focus on bundling it with the computation of disparity or optical flow, leading to a chicken-and-egg problem. In this paper, we leverage convolutional neural network to liberate the occlusion detection task from the interleaved, traditional calculation framework. We propose a Symmetric Network (SymmNet) to directly exploit information from an image pair, without estimating disparity or motion in advance. The proposed network is structurally left-right symmetric to learn the binocular occlusion simultaneously, aimed at jointly improving both results. The comprehensive experiments show that our model achieves state-of-the-art results on detecting the stereo and motion occlusion.

Jie Lyu, Zejian Yuan, Dapeng Chen

For real-world driver drowsiness detection from videos, the variation of head pose is so large that the existing methods on global face is not capable of extracting effective features, such as looking aside and lowering head. Temporal dependencies with variable length are also rarely considered by the previous approaches, e.g., yawning and speaking. In this paper, we propose a Long-term Multi-granularity Deep Framework to detect driver drowsiness in driving videos containing the frontal faces. The framework includes two key components: (1) Multi-granularity Convolutional Neural Network (MCNN), a novel network utilizes a group of parallel CNN extractors on well-aligned facial patches of different granularities, and extracts facial representations effectively for large variation of head pose, furthermore, it can flexibly fuse both detailed appearance clues of the main parts and local to global spatial constraints; (2) a deep Long Short Term Memory network is applied on facial representations to explore long-term relationships with variable length over sequential frames, which is capable to distinguish the states with temporal dependencies, such as blinking and closing eyes. Our approach achieves 90.05% accuracy and about 37 fps speed on the evaluation set of the public NTHU-DDD dataset, which is the state-of-the-art method on driver drowsiness detection. Moreover, we build a new dataset named FI-DDD, which is of higher precision of drowsy locations in temporal dimension.

Jie Lyu, Zejian Yuan, Dapeng Chen

We propose a novel recurrent attentional structure to localize and recognize objects jointly. The network can learn to extract a sequence of local observations with detailed appearance and rough context, instead of sliding windows or convolutions on the entire image. Meanwhile, those observations are fused to complete detection and classification tasks. On training, we present a hybrid loss function to learn the parameters of the multi-task network end-to-end. Particularly, the combination of stochastic and object-awareness strategy, named SA, can select more abundant context and ensure the last fixation close to the object. In addition, we build a real-world dataset to verify the capacity of our method in detecting the object of interest including those small ones. Our method can predict a precise bounding box on an image, and achieve high speed on large images without pooling operations. Experimental results indicate that the proposed method can mine effective context by several local observations. Moreover, the precision and speed are easily improved by changing the number of recurrent steps. Finally, we will open the source code of our proposed approach.

Huaizu Jiang, Zejian Yuan, Ming-Ming Cheng et al.

Salient object detection has been attracting a lot of interest, and recently various heuristic computational models have been designed. In this paper, we formulate saliency map computation as a regression problem. Our method, which is based on multi-level image segmentation, utilizes the supervised learning approach to map the regional feature vector to a saliency score. Saliency scores across multiple levels are finally fused to produce the saliency map. The contributions lie in two-fold. One is that we propose a discriminate regional feature integration approach for salient object detection. Compared with existing heuristic models, our proposed method is able to automatically integrate high-dimensional regional saliency features and choose discriminative ones. The other is that by investigating standard generic region properties as well as two widely studied concepts for salient object detection, i.e., regional contrast and backgroundness, our approach significantly outperforms state-of-the-art methods on six benchmark datasets. Meanwhile, we demonstrate that our method runs as fast as most existing algorithms.