arXiv subjects

Yihang Liu

Publications and source records attributed to Yihang Liu.

31 records · Page 2Linked to original sources

arXiv2025

AFiRe: Anatomy-Driven Self-Supervised Learning for Fine-Grained Representation in Radiographic Images

Current self-supervised methods, such as contrastive learning, predominantly focus on global discrimination, neglecting the critical fine-grained anatomical details required for accurate radiographic analysis. To address this challenge, we propose an Anatomy-driven self-supervised framework for enhancing Fine-grained Representation in radiographic image analysis (AFiRe). The core idea of AFiRe is to align the anatomical consistency with the unique token-processing characteristics of Vision Transformer. Specifically, AFiRe synergistically performs two self-supervised schemes: (i) Token-wise anatomy-guided contrastive learning, which aligns image tokens based on structural and categorical consistency, thereby enhancing fine-grained spatial-anatomical discrimination; (ii) Pixel-level anomaly-removal restoration, which particularly focuses on local anomalies, thereby refining the learned discrimination with detailed geometrical information. Additionally, we propose Synthetic Lesion Mask to enhance anatomical diversity while preserving intra-consistency, which is typically corrupted by traditional data augmentations, such as Cropping and Affine transformations. Experimental results show that AFiRe: (i) provides robust anatomical discrimination, achieving more cohesive feature clusters compared to state-of-the-art contrastive learning methods; (ii) demonstrates superior generalization, surpassing 7 radiography-specific self-supervised methods in multi-label classification tasks with limited labeling; and (iii) integrates fine-grained information, enabling precise anomaly detection using only image-level annotations.

cs.CV↗

arXiv2024

3D-GRES: Generalized 3D Referring Expression Segmentation

3D Referring Expression Segmentation (3D-RES) is dedicated to segmenting a specific instance within a 3D space based on a natural language description. However, current approaches are limited to segmenting a single target, restricting the versatility of the task. To overcome this limitation, we introduce Generalized 3D Referring Expression Segmentation (3D-GRES), which extends the capability to segment any number of instances based on natural language instructions. In addressing this broader task, we propose the Multi-Query Decoupled Interaction Network (MDIN), designed to break down multi-object segmentation tasks into simpler, individual segmentations. MDIN comprises two fundamental components: Text-driven Sparse Queries (TSQ) and Multi-object Decoupling Optimization (MDO). TSQ generates sparse point cloud features distributed over key targets as the initialization for queries. Meanwhile, MDO is tasked with assigning each target in multi-object scenarios to different queries while maintaining their semantic consistency. To adapt to this new task, we build a new dataset, namely Multi3DRes. Our comprehensive evaluations on this dataset demonstrate substantial enhancements over existing models, thus charting a new path for intricate multi-object 3D scene comprehension. The benchmark and code are available at https://github.com/sosppxo/MDIN.

cs.CV↗

arXiv2024

On degree power sum in $P_k$-free graphs

Let $G$ be a graph on $n$ vertices with degree sequence $(d_1,d_2......d_n)$. For a real $p \geq 1$, let $D_p(G)=\sum_{i=1}^nd_i^p$. A Turán-type problem of degree power sum was initiated by Caro and Yuster \cite{caro2000degpower}: determining the function $D_p(n,H) :=\max \{D_p(G): \text{$G$ is an $n$-vertex $H$-free graph}\}$. They obtained some exact values for certain graphs $H$. For a path $P_k$, they mentioned that ``a close examination of the proof of Theorem 1.2 shows that the value of $n_0(k)$ in the statement of the theorem is $O(k^2)$", namely, they could show the $n$-vertex $P_k$-free graph with maximum degree power sum is $W_{n,k-1,\lfloor \frac{k}{2} \rfloor -1} = K_{\lfloor \frac{k}{2} \rfloor -1} \vee \left((n - \lceil \frac{k}{2} \rceil)K_1 \cup K_{1+k-2\lfloor \frac{k}{2} \rfloor} \right)$ when $n \geq c k^2$ for some constant $c$. In this note, we improve their result to a linear size of $k$ by a different approach. The bound is tight up to a constant factor.

math.CO↗

arXiv2024

Solution on strong partition of $2$-balanced regular multipartite tournaments

We call a partition of a $c$-partite tournament into tournaments of order $c$ is strong if each tournament is strongly connected. The strong partition number denoted as $ST(r)$, represents the minimum integer $c'$ such that every regular $r$-balanced $c$-partite tournament has a strong partition with $c\geq c'$. Figueroa, Montellano-Ballesteros and Olsen showed the existence of $ST(r)$ for all $r\geq 2$ and proved that $5\leq ST(2)\leq 7$. In this note, we establish that $ST(2)=6$ and we also show the unique $2$-balanced $5$-partite tournament which has no strong partition.

math.CO↗

arXiv2023

Graph vs. Sequence: An Empirical Study on Knowledge Forms for Knowledge-Grounded Dialogue

Knowledge-grounded dialogue is a task of generating an informative response based on both the dialogue history and external knowledge source. In general, there are two forms of knowledge: manually annotated knowledge graphs and knowledge text from website. From various evaluation viewpoints, each type of knowledge has advantages and downsides. To further distinguish the principles and determinants from the intricate factors, we conduct a thorough experiment and study on the task to answer three essential questions. The questions involve the choice of appropriate knowledge form, the degree of mutual effects between knowledge and the model selection, and the few-shot performance of knowledge. Supported by statistical shreds of evidence, we offer conclusive solutions and sensible suggestions for directions and standards of future research.

cs.CL↗

arXiv2023

FeaInfNet: Diagnosis in Medical Image with Feature-Driven Inference and Visual Explanations

Interpretable deep learning models have received widespread attention in the field of image recognition. Due to the unique multi-instance learning of medical images and the difficulty in identifying decision-making regions, many interpretability models that have been proposed still have problems of insufficient accuracy and interpretability in medical image disease diagnosis. To solve these problems, we propose feature-driven inference network (FeaInfNet). Our first key innovation involves proposing a feature-based network reasoning structure, which is applied to FeaInfNet. The network of this structure compares the similarity of each sub-region image patch with the disease templates and normal templates that may appear in the region, and finally combines the comparison of each sub-region to make the final diagnosis. It simulates the diagnosis process of doctors to make the model interpretable in the reasoning process, while avoiding the misleading caused by the participation of normal areas in reasoning. Secondly, we propose local feature masks (LFM) to extract feature vectors in order to provide global information for these vectors, thus enhancing the expressive ability of the FeaInfNet. Finally, we propose adaptive dynamic masks (Adaptive-DM) to interpret feature vectors and prototypes into human-understandable image patches to provide accurate visual interpretation. We conducted qualitative and quantitative experiments on multiple publicly available medical datasets, including RSNA, iChallenge-PM, Covid-19, ChinaCXRSet, and MontgomerySet. The results of our experiments validate that our method achieves state-of-the-art performance in terms of classification accuracy and interpretability compared to baseline methods in medical image diagnosis. Additional ablation studies verify the effectiveness of each of our proposed components.

cs.CV↗

arXiv2023

Hierarchical Dynamic Masks for Visual Explanation of Neural Networks

Saliency methods generating visual explanatory maps representing the importance of image pixels for model classification is a popular technique for explaining neural network decisions. Hierarchical dynamic masks (HDM), a novel explanatory maps generation method, is proposed in this paper to enhance the granularity and comprehensiveness of saliency maps. First, we suggest the dynamic masks (DM), which enables multiple small-sized benchmark mask vectors to roughly learn the critical information in the image through an optimization method. Then the benchmark mask vectors guide the learning of large-sized auxiliary mask vectors so that their superimposed mask can accurately learn fine-grained pixel importance information and reduce the sensitivity to adversarial perturbations. In addition, we construct the HDM by concatenating DM modules. These DM modules are used to find and fuse the regions of interest in the remaining neural network classification decisions in the mask image in a learning-based way. Since HDM forces DM to perform importance analysis in different areas, it makes the fused saliency map more comprehensive. The proposed method outperformed previous approaches significantly in terms of recognition and localization capabilities when tested on natural and medical datasets.

cs.CV↗

arXiv2022

Decoupling Deep Learning for Interpretable Image Recognition

The interpretability of neural networks has recently received extensive attention. Previous prototype-based explainable networks involved prototype activation in both reasoning and interpretation processes, requiring specific explainable structures for the prototype, thus making the network less accurate as it gains interpretability. Therefore, the decoupling prototypical network (DProtoNet) was proposed to avoid this problem. This new model contains encoder, inference, and interpretation modules. As regards the encoder module, unrestricted feature masks were presented to generate expressive features and prototypes. Regarding the inference module, a multi-image prototype learning method was introduced to update prototypes so that the network can learn generalized prototypes. Finally, concerning the interpretation module, a multiple dynamic masks (MDM) decoder was suggested to explain the neural network, which generates heatmaps using the consistent activation of the original image and mask image at the detection nodes of the network. It decouples the inference and interpretation modules of a prototype-based network by avoiding the use of prototype activation to explain the network's decisions in order to simultaneously improve the accuracy and interpretability of the neural network. The multiple public general and medical datasets were tested, and the results confirmed that our method could achieve a 5% improvement in accuracy and state-of-the-art interpretability compared with previous methods.

cs.CV↗

arXiv2022

Variational Transformer: A Framework Beyond the Trade-off between Accuracy and Diversity for Image Captioning

Accuracy and Diversity are two essential metrizable manifestations in generating natural and semantically correct captions. Many efforts have been made to enhance one of them with another decayed due to the trade-off gap. In this work, we will show that the inferior standard of accuracy draws from human annotations (leave-one-out) are not appropriate for machine-generated captions. To improve diversity with a solid accuracy performance, we exploited a novel Variational Transformer framework. By introducing the "Invisible Information Prior" and the "Auto-selectable GMM", we instruct the encoder to learn the precise language information and object relation in different scenes for accuracy assurance. By introducing the "Range-Median Reward" baseline, we retain more diverse candidates with higher rewards during the RL-based training process for diversity assurance. Experiments show that our method achieves the simultaneous promotion of accuracy (CIDEr) and diversity (self-CIDEr), up to 1.1 and 4.8 percent. Also, our method got the most similar performance of the semantic retrieval compared to human annotations, with 50.3 (50.6 of human) for R@1(i2t).

cs.CV↗

arXiv2022

MDM: Multiple Dynamic Masks for Visual Explanation of Neural Networks

The Class Activation Map (CAM) lookup of a neural network tells us to which regions the neural network focuses when it makes a decision. In the past, the CAM search method was dependent upon a specific internal module of the network. It has specific constraints on the structure of the neural network. To make the search of CAM have generality and high performance. We propose a learning-based algorithm, namely Multiple Dynamic Masks (MDM). It is based on a public cognition that only active features of a picture related to classification will affect the classification results of the neural network, and other features will hardly affect the classification results of the network. The mask generated by MDM conforms to the above cognition. It trains mask vectors of different sizes by constraining mask values and activating consistency, then it uses stacking masks of different scale to generate CAM that can balance spatial information and semantic information. Comparing the results of MDM with those of the recent advanced CAM search method, the performance of MDM has reached the state of the art results. We applied the MDM method to the interpretable neural networks ProtoPNet and XProtoNet, which improved the performance of model in the explainable prototype search. Finally, we visualized the CAM generation effect of MDM on neural networks of different architectures, verifying the generality of the MDM method.

cs.CV↗

arXiv2020

High-speed and high-efficiency three-dimensional shape measurement based on Gray-coded light

Fringe projection profilometry has been increasingly sought and applied in dynamic three-dimensional (3D) shape measurement. In this work, a robust and high-efficiency 3D measurement based on Gray-code light is proposed. Unlike the traditional method, a novel tripartite phase unwrapping method is proposed to avoid the jump errors on the boundary of code words, which are mainly caused by the defocusing of the projector and the motion of the tested object. Subsequently, the time-overlapping coding strategy is presented to greatly increase the coding efficiency, decreasing the projected number in each group, e.g. from 7 (3 + 4) to 4 (3 + 1) for one restored 3D frame. Combination of two proposed techniques allows to reconstruct a pixel-wise and unambiguous 3D geometry of dynamic scenes with strong noise using every 4 projected patterns. The presented techniques preserve the high anti-noise ability of Gray-coded-based method while overcoming the drawbacks of jump errors and low coding efficiency. Experiments have demonstrated that the proposed method can achieve the robust and high-efficiency 3D shape measurement of high-speed dynamic scenes even polluted by strong noise.

eess.IV↗

arXiv2016

High Performance WSe2 Field-Effect Transistors via Controlled Formation of In-Plane Heterojunctions

Monolayer WSe2 is a two dimensional (2D) semiconductor with a direct bandgap, and it has been recently explored as a promising material for electronics and optoelectronics. Low field effect mobility is the main constraint preventing WSe2 from becoming one of the competing channel materials for field-effect transistors (FETs). Recent results have demonstrated that chemical treatments can modify the electrical properties of transition metal dichalcogenides (TMDCs) including MoS2 and WSe2. Here, we report that controlled heating in air significantly improves device performance of WSe2 FETs in terms of on-state currents and field-effect mobilities. Specifically, after heating at optimized conditions, chemical vapor deposition grown monolayer WSe2 FETs showed an average FET mobility of 31 cm2/Vs and on/off current ratios up to 5*108. For few-layer WSe2 FETs, after the same treatment applied, we achieved a high mobility up to 92 cm2/Vs. These values are significantly higher than FETs fabricated using as-grown WSe2 flakes without heating treatment, demonstrating the effectiveness of air heating on the performance improvements of WSe2 FETs. The underlying chemical processes involved during air heating and the formation of in-plane heterojunctions of WSe2 and WO3-x, which is believed to be the reason for the improved FET performance, were studied by spectroscopy and transmission electron microscopy. We further demonstrated that by combining air heating method developed in this work with supporting 2D materials on BN substrate, we achieved a noteworthy field effect mobility of 83 cm2/Vs for monolayer WSe2 FETs. This work is a step towards controlled modification of the properties of WSe2 and potentially other TMDCs, and may greatly improve device performance for future applications of 2D materials in electronics and optoelectronics.

cond-mat.mtrl-sci↗

arXiv2014

Screw-Dislocation-Driven Growth of Two-Dimensional Few-Layer and Pyramid-Like WSe2 by Sulfur-Assisted Chemical Vapor Deposition

Two-dimensional (2D) layered tungsten diselenides (WSe2) material has recently drawn a lot of attention due to its unique optoelectronic properties and ambipolar transport behavior. However, direct chemical vapor deposition (CVD) synthesis of 2D WSe2 is not as straightforward as other 2D materials due to the low reactivity between reactants in WSe2 synthesis. In addition, the growth mechanism of WSe2 in such CVD process remains unclear. Here we report the observation of a screw-dislocation-driven (SDD) spiral growth of 2D WSe2 flakes and pyramid-like structures using a sulfur-assisted CVD method. Few-layer and pyramid-like WSe2 flakes instead of monolayer were synthesized by introducing a small amount of sulfur as a reducer to help the selenization of WO3, which is the precursor of tungsten. Clear observations of steps, helical fringes, and herring-bone contours under atomic force microscope characterization reveal the existence of screw dislocations in the as-grown WSe2. The generation and propagation mechanisms of screw dislocations during the growth of WSe2 were discussed. Back-gated field-effect transistors were made on these 2D WSe2 materials, which show on/off current ratios of 106 and mobility up to 44 cm2/Vs.

cond-mat.mtrl-sci↗