Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks114 from public data
- Perceptual Adversarial Networks for Image-to-Image Transformation393
The perceptual adversarial loss is proposed, which undergoes an adversarial training process between the image transformation network and the discriminative network and can be trained alternately to solve image-to-image transformation tasks.
- Evolutionary Generative Adversarial Networks378
A novel GAN framework called evolutionary GANs (E-GANs) is proposed for stable GAN training and improved generative performance, which overcomes the limitations of an individual adversarial training objective and always preserves the well-performing offspring, contributing to progress in, and the success of GAns.
- 317
- A novel deep learning-based method for damage identification of smart building structures311
A novel method based on deep convolutional neural networks to identify and localise damages of building structures equipped with smart control devices that has outstanding generalisation capacity and higher identification accuracy than other commonly used machine learning methods is proposed.
- RKformer: Runge-Kutta Transformer with Random-Connection Attention for Infrared Small Target Detection259
A novel RKformer model with an encoder-decoder structure, where four specifically designed Runge-Kutta transformer (RKT) blocks are stacked sequentially in the encoder, and a novel random-connection attention (RCA) block, which has a reservoir structure to learn sparse attention via random connections during training.
- 243
- 194
- Could AI Ethical Anxiety, Perceived Ethical Risks and Ethical Awareness About AI Influence University Students’ Use of Generative AI Products? An Ethical Perspective147
Ethical awareness can positively influence BI, and perceived ethical risks can also increase PER, which can help university students better accept and ethically use generative AI products.
- Recent Advances for Quantum Neural Networks in Generative Learning143
This paper interprets these QGLMs, covering quantum circuit Born machines, quantum generative adversarial networks, quantum Boltzmann machines, and quantum variational autoencoders, as the quantum extension of classical generative learning models, and explores their intrinsic relations and their fundamental differences.
- Diff-Font: Diffusion Model for Robust One-Shot Font Generation105
Diff-Font is the first work to utilize a diffusion model for font generation tasks, and outperforms prior font generation methods in both high-fidelity font style replication and the generation of intricate characters.
- 87
- A Systematic Survey of Regularization and Normalization in GANs85
A comprehensive survey on the regularization and normalization techniques from different perspectives of GANs training is conducted and a new taxonomy is proposed based on these objectives, which are summarized at https://github.com/iceli1007/GANs-Regularization-Review.
- Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition69
The first attempt to perform textual reasoning based on visual semantics in this paper is made by paralleling GTR to the language model in a segmentation-based STR baseline, which can effectively exploit the visual-linguistic complementarity via mutual learning.
- 68
- Modeling Image Composition for Complex Scene Generation64
A method that achieves state-of-the-art results on challenging (few-shot) layout-to-image generation tasks by accurately modeling textures, structures and relationships contained in a complex scene by the Transformer with Focal Attention (TwFA).
- 64
- Robust Unpaired Image Dehazing via Density and Depth Decomposition63
A robust self-augmented image dehazing framework for haze generation and removal that can be robustly trained on, not only synthetic indoor datasets, but also real outdoor scenes with remarkable improvement on the real-world image dehazing.
- Tag Disentangled Generative Adversarial Network for Object Image Re-rendering63
A principled Tag Disentangled Generative Adversarial Networks (TD-GAN) for re-rendering new images for the object of interest from a single image of it by specifying multiple scene properties by specifyingmultiple scene properties.
- 58
- HandRefiner: Refining Malformed Hands in Generated Images by Diffusion-based Conditional Inpainting56
This paper introduces a lightweight post-processing solution called HandRefiner, which leverages the hand mesh reconstruction model that consistently adheres to the correct number of fingers and hand shape, while also being capable of fitting the desired hand pose in the generated image.
- MRI-based Alzheimer’s disease prediction via distilling the knowledge in multi-modal data55
This is the first study that attempts to improve an MRI-based prediction model by leveraging extra supervision distilled from multi-modal information, and the proposed multi-instance probabilities demonstrate a superior capability of representing the complicated atrophy distributions.
- Exposure Trajectory Recovery From Motion Blur50
Exposure trajectories are defined, which represent the motion information contained in a blurry image and explain the causes of motion blur, and a novel motion offset estimation framework is proposed to model pixel-wise displacements of the latent sharp image at multiple timepoints.
- 45
- Unified Discrete Diffusion for Simultaneous Vision-Language Generation44
This work unify the discrete diffusion process for multimodal signals by proposing a unified transition matrix and design a mutual attention module with fused embedding layer and a unified objective function to emphasise the inter-modal linkages, which are vital for multi-modality generation.
- 43
- 43
- Eliminating Contextual Prior Bias for Semantic Image Editing via Dual-Cycle Diffusion40
This work presents a novel approach called Dual-Cycle Diffusion, which generates an unbiased mask to guide image editing, and demonstrates the effectiveness of the method, as it significantly improves the D-CLIP score.
- A parallel attention‐augmented bilinear network for early magnetic resonance imaging‐based diagnosis of Alzheimer's disease38
The parallel attention‐augmented bilinear network is proposed, a novel deep learning framework for AD diagnosis based on a 3D convolutional neural network that directly learns both global and local features from sMRI scans without any prior knowledge.
- A Comprehensive Survey on Data-Efficient GANs in Image Generation37
This paper revisits and defines DE-GANs from the perspective of distribution optimization, and proposes a taxonomy, which classifies the existing methods into three categories: Data Selection, GANs Optimization, and Knowledge Sharing.
- Design of BCC/FCC dual-solid solution refractory high-entropy alloys through CALPHAD, machine learning and experimental methods32
This study designed BCC/FCC dual-phase RHEAs through phase diagram calculations and neural network modeling, representing the first successful compositional design of metastable BCC/FCC dual-phase RHEAs.
- Decomposing Semantic Shifts for Composed Image Retrieval30
A Semantic Shift Network (SSN) is proposed that explicitly decomposes the semantic shifts into two steps: from the reference image to the visual prototype and from the visual prototype to the target image.
- ESceme: Vision-and-Language Navigation with Episodic Scene Memory29
This work introduces a mechanism of Episodic Scene memory (ESceme) for VLN that wakes an agent’s memories of past visits when it enters the current scene and provides a simple yet effective implementation of ESceme.
- Experimental Quantum Stochastic Walks Simulating Associative Memory of Hopfield Neural Networks29
A good match rate of the associative memory between the experimental quantum scheme and the expected result for Hopfield neural networks is demonstrated and the ability of quantum simulation for an important feature of a neural network provides a primary but steady step towards photonic artificial intelligence devices for optimization and computation tasks of greatly improved efficiencies.
- MagicFusion: Boosting Text-to-Image Generation Performance by Fusing Diffusion Models28
This work proposes a simple yet effective method called Saliency-aware Noise Blending (SNB), which can empower the fused text-guided diffusion models to achieve more controllable generation by blending the predicted noises of two diffusion models in a saliency-aware manner.
- 27
- 25
- 24
- Multi-Step Denoising Scheduled Sampling: Towards Alleviating Exposure Bias for Diffusion Models20
A multi-step denoising scheduled sampling (MDSS) strategy to alleviate the exposure bias in DDPMs and demonstrates that this approach is more effective in mitigating exposure bias in DDPM, DDIM, and DPM-solver.
- 20
- 19
- 18
- Null-text Guidance in Diffusion Models is Secretly a Cartoon-style Creator17
It is demonstrated that null-text guidance in diffusion models is secretly a cartoon-style creator, i.e., the generated images can be efficiently transformed into cartoons by simply perturbing the null- Text guidance in the sampling process.
- 17
- 15
- 15
- Effect of Yb^3+ on the upconversion luminescence of Tm^3+/Yb^3+ co-doped La_2O_3-TiO_2-ZrO_2 glasses15
- A novel valence-balanced double half-Heusler Ti2Zr2Hf2NbVFe5Ni3Sb8 alloy by high entropy engineering14
- When ControlNet Meets Inexplicit Masks: A Case Study of ControlNet on its Contour-following Ability12
To enhance controllability with inexplicit masks, an advanced Shape-aware ControlNet consisting of a deterioration estimator and a shape-prior modulation block is devised.
- 12
- 11
- Peer Is Your Pillar: A Data-Unbalanced Conditional GANs for Few-Shot Image Generation11
This work proposes a novel pipeline called Peer is your Pillar (PIP), which combines a target few-shot dataset with a peer dataset to create a data-unbalanced conditional generation and uses a direction loss based on pre-trained CLIP to improve image diversity.
- Neural Maximum a Posteriori Estimation on Unpaired Data for Motion Deblurring11
The proposed NeruMAP consists of a motion estimation network and a deblurring network which are trained jointly to model the (re)blurring process and is the first framework that enables training imagedeblurring networks on unpaired datasets.
- Point Cloud Generation From Multiple Angles of Voxel Grids11
The experimental result of the algorithm indicates that the algorithm is capable of generating high-quality point clouds by understanding the semantic features of the voxel grid.
- 9
- Deep Representation Calibrated Bayesian Neural Network for Semantically Explainable Face Inpainting and Editing9
This work proposes a novel deep representation calibrated Bayesian neural network (DRCBNN) for semantically explainable face inpainting and editing and exploits deep representation into Bayesian decision theory and derive a deep representation calibration evidence lower bound (ELBO).
- 8
- 8
- A fluorescent probe for detecting carboxylesterase level on investigating Chinese medicine water decoction efficacy7
A fluorescent probe ImPI-CE with the imidazo-pyridin-isovanillin sub-structure was developed and studied for monitoring the Carboxylesterase level in colorectal cells on evaluating the Chinese medicine water decoction efficacy and achieved the stable monitoring of multiple Chinese medicine water decoction samples.
- 7
- MMoT: Mixture-of-Modality-Tokens Transformer for Composed Multimodal Conditional Image Synthesis7
A Mixture-of-Modality-Tokens Transformer (MMoT) that adaptively fuses fine-grained multimodal control signals, a multimodal balanced training loss to stabilize the optimization of each modality, and a multimodal sampling guidance to balance the strength of each modality control signal is introduced.
- Decompose Semantic Shifts for Composed Image Retrieval7
A Semantic Shift network (SSN) is proposed that explicitly decomposes the semantic shifts into two steps: from the reference image to the visual prototype and from theVisual prototype to the target image.
- Leveraging GAN Priors for Few-Shot Part Segmentation7
This work designs a fine-tuning strategy to progressively tune an image generator into a segmentation generator, where the supervision of the generator varying from images to segmentation maps by interpolation, and proposes a two-stream architecture, i.e., a segmentations stream to generate task-specific features, and an image stream to provide spatial constraints.
- 7
- Improving Chinese Sentence Polarity Classification via Opinion Paraphrasing7
This paper proposes to improve a supported vector mach ines based polarity classifier by enriching both training data and test data via opinion paraphrasing, and shows that using opinions paraphrases results in significant performance in polarity classification.
- 6
- 6
- 6
- One-Shot Generative Domain Adaptation in 3D GANs6
3D-Adapter is introduced, the first one-shot 3D GDA method, for diverse and faithful generation, and seamlessly extends its capabilities to zero-shot scenarios, and preserves the potential for crucial tasks such as interpolation, reconstruction, and editing within the latent space of the pre-trained generator.
- Cross-modal & Cross-domain Learning for Unsupervised LiDAR Semantic Segmentation6
The experimental results on several datasets show that in the proposed setting, the developed CoMoDaL can achieve segmentation without the supervision of labeled LiDAR data.
- D 2 Animator: Dual Distillation of StyleGAN For High-Resolution Face Animation6
This work devise a dual distillation strategy (D2Animator) for generating animated high-resolution face videos conditioned on identities and poses from different images and proposes an Augmentation-based Distiller that learns to encode arbitrary face deformation into a combination of interpolation directions via training on augmentation samples synthesized by CluDistiller.
- Fabrication and up-conversion fluorescence property of Er3+/Yb3+co-doped Ca-Si-Ti biomaterials6
This work demonstrates bulk-type up-conversion biomaterials which could be used as a bone repair material with the ability to monitor bone mineralization and indicates that Er3+/Yb3+ co-doped CST3 materials are bioactive, in which the HAP mineralization in bone repair could be monitored by measuring the intensity change of up- Conversion fluorescence.
- 5
- 5
- 5
- 5
- 5
- Visual Superordinate Abstraction for Robust Concept Learning4
A visual superordinates abstraction framework for explicitly modeling semantic-aware visual subspaces and a quasi-center visual concept clustering and superordinate shortcut learning schemes are proposed to enhance the discrimination and independence of concepts within each visual superordinate.
- Semantically-Consistent Dynamic Blurry Image Generation for Image Deblurring4
A novel method to generate semantic-aware dense dynamic motion, and employ the generated motion to synthesize blurry images that can generate diverse dynamic motion and visually realistic blurry images is proposed.
- 4
- 3
- 3
- 3
- Digital Media Empower The Marketing Promotion of Health Management: A Case Study of KEEP APP3
According to the findings of this study, in the four measurement dimensions chosen, perceived usefulness, outcome expectations, and social interactions have positive and significant effects on the user's intention to continue using, whereas subjective norms have no influence on the users' continuous using intention.
- Towards High-Fidelity Face Normal Estimation3
A coarse-to-fine framework to estimate face normal from an in-the-wild image with only a coarse exemplar reference is proposed, which can largely alleviate the negative impact caused by lacking training data, and focus on exploring the high-fidelity normal contained in natural images.
- 3
- Self-supervised Exposure Trajectory Recovery for Dynamic Blur Estimation.3
Under mild constraints, the learned motion offsets can recover dense, (non-)linear exposure trajectories, which significantly reduce temporal disorder and ill-posed problems and further contribute to motion-aware image deblurring and warping-based video extraction from a single blurry image.
- 3
- 2
- Towards Modality-agnostic Label-efficient Segmentation with Entropy-Regularized Distribution Alignment2
This work proposes a novel learning strategy to regularize the pseudo-labels generated for training, thus effectively narrowing the gaps between pseudo-labels and model predictions and introduces an Entropy Regularization loss and a Distribution Alignment loss for label-efficient learning, resulting in an ERDA learning strategy.
- 2
- 1
- An Imidazo-Pyridin-Derived Fluorescent Probe for Monitoring Hydrogen Sulfide in Chinese Medicinal Materials1
An imidazo-pyridin-derived fluorescent probe ImPy-HS was developed for monitoring the hydrogen sulfide (H2S) level in the water decoction of Chinese medicinal materials and showed the advantages including relatively high sensitivity, high selectivity, and high steadiness.
- A fluorescent probe with a vanillin–pyridine–imidazole core structure for carboxylesterase detection in macrophage polarization during bone homeostasis1
Tests in solution showed the advantages of VanPI-CarE, including high sensitivity, excellent stability under various working conditions, high selectivity, and low cytotoxicity, and the probe revealed macrophage polarization in bone homeostasis under both induction and inhibition conditions via CarE detection.
- 1
- Dual Mapping of 2D StyleGAN for 3D-Aware Image Generation and Manipulation (Student Abstract)1
A dual-mapping framework to make the generated images of pretrained 2D StyleGAN consistent in 3D space consistent in 3D space and outperforms the 3D-aware GANs in controllability and image quality.
- 1
- 1
- –
- –
- MUG-V 10B: High-efficiency Training Pipeline for Large Video Generation Models–
This is the first public release of large-scale video generation training code that exploits Megatron-Core to achieve high training efficiency and near-linear multi-node scaling, and the complete stack is open-source, including model weights, Megatron-Core-based large-scale training code, and inference pipelines for video generation and enhancement.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Generative modelling and adversarial learning–
This thesis aims to both improve the quality of generative modelling and manipulate generated samples by specifying multiple scene properties and devise a novel model, called a perceptual adversarial network (PAN), which consists of two feed-forward convolutional neural networks: a transformation network and a discriminative network.
- –
- Chinese Tweets Segmentation based on Morphemes–
This paper takes morpheme as the basic component units of Chinese words and investigates the relationship between Chinese new words and their internal morphological structures, and explores the incorporation of multiple features, especially the word -internal morphological featu res, of great value to Chinese tweets segmentation.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.