arXiv subjects

Yanxi Liu

Publications and source records attributed to Yanxi Liu.

13 recordsLinked to original sources

arXiv2026

Perception-and-action system for humanoid robot task execution in construction

Humanoid robots, with their human-like shape and multi-tasking capabilities, are well-aligned with human-dominated workplaces, like those in civil and construction engineering, where they could collaborate with human workers or autonomously perform physically demanding and hazardous tasks. Despite this promise, limited research has explored how to endow these robots with the practical capabilities needed to perform construction tasks. To this end, this study proposes a novel perception-and-action system that enables humanoid robots to learn and perform construction tasks from worker demonstrations. This system contains two deep networks: Humanoid-PoseNet, which extracts human postures and translates them into mechanically feasible poses for a humanoid robot; and Humanoid-ActionNet, which learns robot-executable actions based on these translated poses. Experimental results demonstrate that the humanoid robot reliably executed eight construction-related actions, achieving an average motion-tracking error of 82.45 mm MPJPE (Mean Per Joint Position Error). This work provides an early step toward deploying humanoid collaborators in construction.

cs.RO↗

arXiv2025

FootFormer: Estimating Stability from Visual Input

We propose FootFormer, a cross-modality approach for jointly predicting human motion dynamics directly from visual input. On multiple datasets, FootFormer achieves statistically significantly better or equivalent estimates of foot pressure distributions, foot contact maps, and center of mass (CoM), as compared with existing methods that generate one or two of those measures. Furthermore, FootFormer achieves SOTA performance in estimating stability-predictive components (CoP, CoM, BoS) used in classic kinesiology metrics. Code and data are available at https://github.com/keatonkraiger/Vision-to-Stability.git.

cs.CV↗

arXiv2025

MedReasoner: Reinforcement Learning Drives Reasoning Grounding from Clinical Thought to Pixel-Level Precision

Accurately grounding regions of interest (ROIs) is critical for diagnosis and treatment planning in medical imaging. While multimodal large language models (MLLMs) combine visual perception with natural language, current medical-grounding pipelines still rely on supervised fine-tuning with explicit spatial hints, making them ill-equipped to handle the implicit queries common in clinical practice. This work makes three core contributions. We first define Unified Medical Reasoning Grounding (UMRG), a novel vision-language task that demands clinical reasoning and pixel-level grounding. Second, we release U-MRG-14K, a dataset of 14K samples featuring pixel-level masks alongside implicit clinical queries and reasoning traces, spanning 10 modalities, 15 super-categories, and 108 specific categories. Finally, we introduce MedReasoner, a modular framework that distinctly separates reasoning from segmentation: an MLLM reasoner is optimized with reinforcement learning, while a frozen segmentation expert converts spatial prompts into masks, with alignment achieved through format and accuracy rewards. MedReasoner achieves state-of-the-art performance on U-MRG-14K and demonstrates strong generalization to unseen clinical queries, underscoring the significant promise of reinforcement learning for interpretable medical grounding.

cs.CV↗

arXiv2024

Recurrence-based Vanishing Point Detection

Classical approaches to Vanishing Point Detection (VPD) rely solely on the presence of explicit straight lines in images, while recent supervised deep learning approaches need labeled datasets for training. We propose an alternative unsupervised approach: Recurrence-based Vanishing Point Detection (R-VPD) that uses implicit lines discovered from recurring correspondences in addition to explicit lines. Furthermore, we contribute two Recurring-Pattern-for-Vanishing-Point (RPVP) datasets: 1) a Synthetic Image dataset with 3,200 ground truth vanishing points and camera parameters, and 2) a Real-World Image dataset with 1,400 human annotated vanishing points. We compare our method with two classical methods and two state-of-the-art deep learning-based VPD methods. We demonstrate that our unsupervised approach outperforms all the methods on the synthetic images dataset, outperforms the classical methods, and is on par with the supervised learning approaches on real-world images.

cs.CV↗

arXiv2023

Information-based Optimal Subdata Selection for Clusterwise Linear Regression

Mixture-of-Experts models are commonly used when there exist distinct clusters with different relationships between the independent and dependent variables. Fitting such models for large datasets, however, is computationally virtually impossible. An attractive alternative is to use a subdata selected by ``maximizing" the Fisher information matrix. A major challenge is that no closed-form expression for the Fisher information matrix is available for such models. Focusing on clusterwise linear regression models, a subclass of MoE models, we develop a framework that overcomes this challenge. We prove that the proposed subdata selection approach is asymptotically optimal, i.e., no other method is statistically more efficient than the proposed one when the full data size is large.

stat.ME↗

arXiv2023

Deep fused flow and topology features for botnet detection basing on pretrained GCN

Nowadays, botnets have become one of the major threats to cyber security. The characteristics of botnets are mainly reflected in bots network behavior and their intercommunication relationships. Existing botnet detection methods use flow features or topology features individually, which overlook the other type of feature. This affects model performance. In this paper, we propose a botnet detection model which uses graph convolutional network (GCN) to deeply fuse flow features and topology features for the first time. We construct communication graphs from network traffic and represent nodes with flow features. Due to the imbalance of existing public traffic flow datasets, it is impossible to train a GCN model on these datasets. Therefore, we use a balanced public communication graph dataset to pretrain a GCN model, thereby guaranteeing its capacity for identify topology features. We then feed the communication graph with flow features into the pretrained GCN. The output from the last hidden layer is treated as the fusion of flow and topology features. Additionally, by adjusting the number of layers in the GCN network, the model can effectively detect botnets under both C2 and P2P structures. Validated on the public ISCX2014 dataset, our approach achieves a remarkable recall rate 92.90% and F1-score 92.76% for C2 botnets, alongside recall rate 94.66% and F1-score of 92.35% for P2P botnets. These results not only demonstrate the effectiveness of our method, but also outperform the performance of the currently leading detection models.

cs.CR↗

arXiv2023

EscherNet 101

A deep learning model, EscherNet 101, is constructed to categorize images of 2D periodic patterns into their respective 17 wallpaper groups. Beyond evaluating EscherNet 101 performance by classification rates, at a micro-level we investigate the filters learned at different layers in the network, capable of capturing second-order invariants beyond edge and curvature.

cs.CV↗

arXiv2023

Stock Market Prediction via Deep Learning Techniques: A Survey

Existing surveys on stock market prediction often focus on traditional machine learning methods instead of deep learning methods. This motivates us to provide a structured and comprehensive overview of the research on stock market prediction. We present four elaborated subtasks of stock market prediction and propose a novel taxonomy to summarize the state-of-the-art models based on deep neural networks. In addition, we also provide detailed statistics on the datasets and evaluation metrics commonly used in the stock market. Finally, we point out several future directions by sharing some new perspectives on stock market prediction.

q-fin.GN↗

arXiv2022

Image-based Stability Quantification

Quantitative evaluation of human stability using foot pressure/force measurement hardware and motion capture (mocap) technology is expensive, time consuming, and restricted to the laboratory. We propose a novel image-based method to estimate three key components for stability computation: Center of Mass (CoM), Base of Support (BoS), and Center of Pressure (CoP). Furthermore, we quantitatively validate our image-based methods for computing two classic stability measures, CoMtoCoP and CoMtoBoS distances, against values generated directly from laboratory-based sensor output (ground truth) using a publicly available, multi-modality (mocap, foot pressure, two-view videos), ten-subject human motion dataset. Using Leave One Subject Out (LOSO) cross-validation, experimental results show: 1) our image-based CoM estimation method (CoMNet) consistently outperforms state-of-the-art inertial sensor-based CoM estimation techniques; 2) stability computed by our image-based method combined with insole foot pressure sensor data produces consistent, strong, and statistically significant correlation with ground truth stability measures (CoMtoCoP r = 0.79 p < 0.001, CoMtoBoS r = 0.75 p < 0.001); 3) our fully image-based estimation of stability produces consistent, positive, and statistically significant correlation on the two stability metrics (CoMtoCoP r = 0.31 p < 0.001, CoMtoBoS r = 0.22 p < 0.043). Our study provides promising quantitative evidence for the feasibility of image-based stability evaluation in natural environments.

cs.CV↗

arXiv2022

Novel 3D Scene Understanding Applications From Recurrence in a Single Image

We demonstrate the utility of recurring pattern discovery from a single image for spatial understanding of a 3D scene in terms of (1) vanishing point detection, (2) hypothesizing 3D translation symmetry and (3) counting the number of RP instances in the image. Furthermore, we illustrate the feasibility of leveraging RP discovery output to form a more precise, quantitative text description of the scene. Our quantitative evaluations on a new 1K+ Recurring Pattern (RP) benchmark with diverse variations show that visual perception of recurrence from one single view leads to scene understanding outcomes that are as good as or better than existing supervised methods and/or unsupervised methods that use millions of images.

cs.CV↗

arXiv2020

From Kinematics To Dynamics: Estimating Center of Pressure and Base of Support from Video Frames of Human Motion

To gain an understanding of the relation between a given human pose image and the corresponding physical foot pressure of the human subject, we propose and validate two end-to-end deep learning architectures, PressNet and PressNet-Simple, to regress foot pressure heatmaps (dynamics) from 2D human pose (kinematics) derived from a video frame. A unique video and foot pressure data set of 813,050 synchronized pairs, composed of 5-minute long choreographed Taiji movement sequences of 6 subjects, is collected and used for leaving-one-subject-out cross validation. Our initial experimental results demonstrate reliable and repeatable foot pressure prediction from a single image, setting the first baseline for such a complex cross modality mapping problem in computer vision. Furthermore, we compute and quantitatively validate the Center of Pressure (CoP) and Base of Support (BoS) from predicted foot pressure distribution, obtaining key components in pose stability analysis from images with potential applications in kinesiology, medicine, sports and robotics.

cs.CV↗

arXiv2019

Learning Dynamics from Kinematics: Estimating 2D Foot Pressure Maps from Video Frames

Pose stability analysis is the key to understanding locomotion and control of body equilibrium, with applications in numerous fields such as kinesiology, medicine, and robotics. In biomechanics, Center of Pressure (CoP) is used in studies of human postural control and gait. We propose and validate a novel approach to learn CoP from pose of a human body to aid stability analysis. More specifically, we propose an end-to-end deep learning architecture to regress foot pressure heatmaps, and hence the CoP locations, from 2D human pose derived from video. We have collected a set of long (5min +) choreographed Taiji (Tai Chi) sequences of multiple subjects with synchronized foot pressure and video data. The derived human pose data and corresponding foot pressure maps are used jointly in training a convolutional neural network with residual architecture, named PressNET. Cross-subject validation results show promising performance of PressNET, significantly outperforming the baseline method of K-Nearest Neighbors. Furthermore, we demonstrate that our computation of center of pressure (CoP) from PressNET is not only significantly more accurate than those obtained from the baseline approach but also meets the expectations of corresponding lab-based measurements of stability studies in kinesiology.

cs.CV↗

arXiv2017

Beyond Planar Symmetry: Modeling human perception of reflection and rotation symmetries in the wild

Humans take advantage of real world symmetries for various tasks, yet capturing their superb symmetry perception mechanism with a computational model remains elusive. Motivated by a new study demonstrating the extremely high inter-person accuracy of human perceived symmetries in the wild, we have constructed the first deep-learning neural network for reflection and rotation symmetry detection (Sym-NET), trained on photos from MS-COCO (Microsoft-Common Object in COntext) dataset with nearly 11K consistent symmetry-labels from more than 400 human observers. We employ novel methods to convert discrete human labels into symmetry heatmaps, capture symmetry densely in an image and quantitatively evaluate Sym-NET against multiple existing computer vision algorithms. On CVPR 2013 symmetry competition testsets and unseen MS-COCO photos, Sym-NET significantly outperforms all other competitors. Beyond mathematically well-defined symmetries on a plane, Sym-NET demonstrates abilities to identify viewpoint-varied 3D symmetries, partially occluded symmetrical objects, and symmetries at a semantic level.

cs.CV↗