arXiv subjects

Yang Zhao

Publications and source records attributed to Yang Zhao.

At least 271 records · Page 15Linked to original sources

arXiv2023

HiFi Tuner: High-Fidelity Subject-Driven Fine-Tuning for Diffusion Models

This paper explores advancements in high-fidelity personalized image generation through the utilization of pre-trained text-to-image diffusion models. While previous approaches have made significant strides in generating versatile scenes based on text descriptions and a few input images, challenges persist in maintaining the subject fidelity within the generated images. In this work, we introduce an innovative algorithm named HiFi Tuner to enhance the appearance preservation of objects during personalized image generation. Our proposed method employs a parameter-efficient fine-tuning framework, comprising a denoising process and a pivotal inversion process. Key enhancements include the utilization of mask guidance, a novel parameter regularization technique, and the incorporation of step-wise subject representations to elevate the sample fidelity. Additionally, we propose a reference-guided generation approach that leverages the pivotal inversion of a reference image to mitigate unwanted subject variations and artifacts. We further extend our method to a novel image editing task: substituting the subject in an image through textual manipulations. Experimental evaluations conducted on the DreamBooth dataset using the Stable Diffusion model showcase promising results. Fine-tuning solely on textual embeddings improves CLIP-T score by 3.6 points and improves DINO score by 9.6 points over Textual Inversion. When fine-tuning all parameters, HiFi Tuner improves CLIP-T score by 1.2 points and improves DINO score by 1.2 points over DreamBooth, establishing a new state of the art.

cs.CV↗

arXiv2023

Role Taxonomy of Units in Deep Neural Networks

Identifying the role of network units in deep neural networks (DNNs) is critical in many aspects including giving understandings on the mechanisms of DNNs and building basic connections between deep learning and neuroscience. However, there remains unclear on which roles the units in DNNs with different generalization ability could present. To this end, we give role taxonomy of units in DNNs via introducing the retrieval-of-function test, where units are categorized into four types in terms of their functional preference on separately the training set and testing set. We show that ratios of the four categories are highly associated with the generalization ability of DNNs from two distinct perspectives, based on which we give signs of DNNs with well generalization.

cs.CV↗

arXiv2023

Estimating the Generalization in Deep Neural Networks via Sparsity

Generalization is the key capability for deep neural networks (DNNs). However, it is challenging to give a reliable measure of the generalization ability of a DNN via only its nature. In this paper, we propose a novel method for estimating the generalization gap based on network sparsity. In our method, two key quantities are proposed first. They have close relationship with the generalization ability and can be calculated directly from the training results alone. Then a simple linear model involving two key quantities are constructed to give accurate estimation of the generalization gap. By training DNNs with a wide range of generalization gap on popular datasets, we show that our key quantities and linear model could be efficient tools for estimating the generalization gap of DNNs.

cs.CV↗

arXiv2023

Cut-and-Paste: Subject-Driven Video Editing with Attention Control

This paper presents a novel framework termed Cut-and-Paste for real-word semantic video editing under the guidance of text prompt and additional reference image. While the text-driven video editing has demonstrated remarkable ability to generate highly diverse videos following given text prompts, the fine-grained semantic edits are hard to control by plain textual prompt only in terms of object details and edited region, and cumbersome long text descriptions are usually needed for the task. We therefore investigate subject-driven video editing for more precise control of both edited regions and background preservation, and fine-grained semantic generation. We achieve this goal by introducing an reference image as supplementary input to the text-driven video editing, which avoids racking your brain to come up with a cumbersome text prompt describing the detailed appearance of the object. To limit the editing area, we refer to a method of cross attention control in image editing and successfully extend it to video editing by fusing the attention map of adjacent frames, which strikes a balance between maintaining video background and spatio-temporal consistency. Compared with current methods, the whole process of our method is like ``cut" the source object to be edited and then ``paste" the target object provided by reference image. We demonstrate that our method performs favorably over prior arts for video editing under the guidance of text prompt and extra reference image, as measured by both quantitative and subjective evaluations.

cs.CV↗

arXiv2023

Dissipative Landau-Zener transitions in a three-level bow-tie model: accurate dynamics with the Davydov multi-D2 Ansatz

We investigate Landau-Zener (LZ) transitions in the three-level bow-tie model (3L-BTM) in a dissipative environment by using the numerically accurate method of multiple Davydov D2 Ansatze. We first consider the 3L-TBM coupled to a single harmonic mode, study evolutions of the transition probabilities for selected values of the model parameters, and interpret the obtained results with the aid of the energy diagram method. We then explore the 3L-TBM coupled to a boson bath. Our simulations demonstrate that sub-Ohmic, Ohmic and super-Ohmic boson baths have substantially different influences on the 3L-BTM dynamics, which cannot be grasped by the standard phenomenological Markovian single-rate descriptions. We also describe novel bath-induced phenomena which are absent in two-level LZ systems.

quant-ph↗

arXiv2023

Intell-dragonfly: A Cybersecurity Attack Surface Generation Engine Based On Artificial Intelligence-generated Content Technology

With the rapid development of the Internet, cyber security issues have become increasingly prominent. Traditional cyber security defense methods are limited in the face of ever-changing threats, so it is critical to seek innovative attack surface generation methods. This study proposes Intell-dragonfly, a cyber security attack surface generation engine based on artificial intelligence generation technology, to meet the challenges of cyber security. Based on ChatGPT technology, this paper designs an automated attack surface generation process, which can generate diversified and personalized attack scenarios, targets, elements and schemes. Through experiments in a real network environment, the effect of the engine is verified and compared with traditional methods, which improves the authenticity and applicability of the attack surface. The experimental results show that the ChatGPT-based method has significant advantages in the accuracy, diversity and operability of attack surface generation. Furthermore, we explore the strengths and limitations of the engine and discuss its potential applications in the field of cyber security. This research provides a novel approach to the field of cyber security that is expected to have a positive impact on defense and prevention of cyberthreats.

cs.CR↗

arXiv2023

RIScatter: Unifying Backscatter Communication and Reconfigurable Intelligent Surface

Backscatter Communication (BackCom) nodes harvest energy from and modulate information over external carriers. Reconfigurable Intelligent Surface (RIS) adapts phase shift response to alter channel strength in specific directions. In this paper, we unify those two seemingly different technologies (and their derivatives) into one architecture called RIScatter. RIScatter is a batteryless cognitive radio that recycles ambient signal in an adaptive and customizable manner, where dispersed or co-located scatter nodes partially modulate their information and partially engineer the wireless channel. The key is to render the probability distribution of reflection states as a joint function of the information source, Channel State Information (CSI), and relative priority of coexisting links. This enables RIScatter to softly bridge BackCom and RIS; reduce to either in special cases; or evolve in a mixed form for heterogeneous traffic control and universal hardware design. We also propose a low-complexity Successive Interference Cancellation (SIC)-free receiver that exploits the properties of RIScatter. For a single-user multi-node network, we characterize the achievable primary-(total-)backscatter rate region by optimizing the input distribution at scatter nodes, the active beamforming at the Access Point (AP), and the energy decision regions at the user. Simulations demonstrate RIScatter nodes can shift between backscatter modulation and passive beamforming.

cs.IT↗

arXiv2023

Connecting Multi-modal Contrastive Representations

Multi-modal Contrastive Representation learning aims to encode different modalities into a semantically aligned shared space. This paradigm shows remarkable generalization ability on numerous downstream tasks across various modalities. However, the reliance on massive high-quality data pairs limits its further development on more modalities. This paper proposes a novel training-efficient method for learning MCR without paired data called Connecting Multi-modal Contrastive Representations (C-MCR). Specifically, given two existing MCRs pre-trained on (A, B) and (B, C) modality pairs, we project them to a new space and use the data from the overlapping modality B to aligning the two MCRs in the new space. Meanwhile, since the modality pairs (A, B) and (B, C) are already aligned within each MCR, the connection learned by overlapping modality can also be transferred to non-overlapping modality pair (A, C). To unleash the potential of C-MCR, we further introduce a semantic-enhanced inter- and intra-MCR connection method. We first enhance the semantic consistency and completion of embeddings across different modalities for more robust alignment. Then we utilize the inter-MCR alignment to establish the connection, and employ the intra-MCR alignment to better maintain the connection for inputs from non-overlapping modalities. To demonstrate the effectiveness of C-MCR, we connect CLIP and CLAP via texts to derive audio-visual representations, and integrate CLIP and ULIP via images for 3D-language representations. Remarkably, without using any paired data, C-MCR for audio-visual achieves state-of-the-art performance on audio-image retrieval, audio-visual source localization, and counterfactual audio-image recognition tasks. Furthermore, C-MCR for 3D-language also attains advanced zero-shot 3D point cloud classification accuracy on ModelNet40.

cs.LG↗

arXiv2023

Extending Multi-modal Contrastive Representations

Multi-modal contrastive representation (MCR) of more than three modalities is critical in multi-modal learning. Although recent methods showcase impressive achievements, the high dependence on large-scale, high-quality paired data and the expensive training costs limit their further development. Inspired by recent C-MCR, this paper proposes Extending Multimodal Contrastive Representation (Ex-MCR), a training-efficient and paired-data-free method to flexibly learn unified contrastive representation space for more than three modalities by integrating the knowledge of existing MCR spaces. Specifically, Ex-MCR aligns multiple existing MCRs into the same based MCR, which can effectively preserve the original semantic alignment of the based MCR. Besides, we comprehensively enhance the entire learning pipeline for aligning MCR spaces from the perspectives of training data, architecture, and learning objectives. With the preserved original modality alignment and the enhanced space alignment, Ex-MCR shows superior representation learning performance and excellent modality extensibility. To demonstrate the effectiveness of Ex-MCR, we align the MCR spaces of CLAP (audio-text) and ULIP (3D-vision) into the CLIP (vision-text), leveraging the overlapping text and image modality, respectively. Remarkably, without using any paired data, Ex-MCR learns a 3D-image-text-audio unified contrastive representation, and it achieves state-of-the-art performance on audio-visual, 3D-image, audio-text, visual-text retrieval, and 3D object classification tasks. More importantly, extensive qualitative results further demonstrate the emergent semantic alignment between the extended modalities (e.g., audio and 3D), which highlights the great potential of modality extensibility.

cs.CV↗

arXiv2023

Cross-Dataset-Robust Method for Blind Real-World Image Quality Assessment

Although many effective models and real-world datasets have been presented for blind image quality assessment (BIQA), recent BIQA models usually tend to fit specific training set. Hence, it is still difficult to accurately and robustly measure the visual quality of an arbitrary real-world image. In this paper, a robust BIQA method, is designed based on three aspects, i.e., robust training strategy, large-scale real-world dataset, and powerful backbone. First, many individual models based on popular and state-of-the-art (SOTA) Swin-Transformer (SwinT) are trained on different real-world BIQA datasets respectively. Then, these biased SwinT-based models are jointly used to generate pseudo-labels, which adopts the probability of relative quality of two random images instead of fixed quality score. A large-scale real-world image dataset with 1,000,000 image pairs and pseudo-labels is then proposed for training the final cross-dataset-robust model. Experimental results on cross-dataset tests show that the performance of the proposed method is even better than some SOTA methods that are directly trained on these datasets, thus verifying the robustness and generalization of our method.

cs.CV↗

arXiv2023

BenchTemp: A General Benchmark for Evaluating Temporal Graph Neural Networks

To handle graphs in which features or connectivities are evolving over time, a series of temporal graph neural networks (TGNNs) have been proposed. Despite the success of these TGNNs, the previous TGNN evaluations reveal several limitations regarding four critical issues: 1) inconsistent datasets, 2) inconsistent evaluation pipelines, 3) lacking workload diversity, and 4) lacking efficient comparison. Overall, there lacks an empirical study that puts TGNN models onto the same ground and compares them comprehensively. To this end, we propose BenchTemp, a general benchmark for evaluating TGNN models on various workloads. BenchTemp provides a set of benchmark datasets so that different TGNN models can be fairly compared. Further, BenchTemp engineers a standard pipeline that unifies the TGNN evaluation. With BenchTemp, we extensively compare the representative TGNN models on different tasks (e.g., link prediction and node classification) and settings (transductive and inductive), w.r.t. both effectiveness and efficiency metrics. We have made BenchTemp publicly available at https://github.com/qianghuangwhu/benchtemp.

cs.LG↗

arXiv2023

Chat-3D: Data-efficiently Tuning Large Language Model for Universal Dialogue of 3D Scenes

3D scene understanding has gained significant attention due to its wide range of applications. However, existing methods for 3D scene understanding are limited to specific downstream tasks, which hinders their practicality in real-world applications. This paper presents Chat-3D, which combines the 3D visual perceptual ability of pre-trained 3D representations and the impressive reasoning and conversation capabilities of advanced LLMs to achieve the first universal dialogue systems for 3D scenes. Specifically, we align 3D representations into the feature space of LLMs, thus enabling LLMs to perceive the 3D world. Given the scarcity of 3D scene-text data, we propose a three-stage training strategy to efficiently utilize the available data for better alignment. To enhance the reasoning ability and develop a user-friendly interaction scheme, we further construct a high-quality object-centric 3D instruction dataset and design an associated object-centric prompt. Our experiments show that Chat-3D achieves an impressive ability to comprehend diverse instructions for 3D scenes, engage in intricate spatial reasoning, and incorporate external knowledge into its responses. Chat-3D achieves a 75.6% relative score compared with GPT-4 on the constructed instruction dataset.

cs.CV↗

arXiv2023

RPG-Palm: Realistic Pseudo-data Generation for Palmprint Recognition

Palmprint recently shows great potential in recognition applications as it is a privacy-friendly and stable biometric. However, the lack of large-scale public palmprint datasets limits further research and development of palmprint recognition. In this paper, we propose a novel realistic pseudo-palmprint generation (RPG) model to synthesize palmprints with massive identities. We first introduce a conditional modulation generator to improve the intra-class diversity. Then an identity-aware loss is proposed to ensure identity consistency against unpaired training. We further improve the Bézier palm creases generation strategy to guarantee identity independence. Extensive experimental results demonstrate that synthetic pretraining significantly boosts the recognition model performance. For example, our model improves the state-of-the-art BézierPalm by more than $5\%$ and $14\%$ in terms of TAR@FAR=1e-6 under the $1:1$ and $1:3$ Open-set protocol. When accessing only $10\%$ of the real training data, our method still outperforms ArcFace with $100\%$ real training data, indicating that we are closer to real-data-free palmprint recognition.

cs.CV↗

arXiv2023

Target-point Attention Transformer: A novel trajectory predict network for end-to-end autonomous driving

In the field of autonomous driving, there have been many excellent perception models for object detection, semantic segmentation, and other tasks, but how can we effectively use the perception models for vehicle planning? Traditional autonomous vehicle trajectory prediction methods not only need to obey traffic rules to avoid collisions, but also need to follow the prescribed route to reach the destination. In this paper, we propose a Transformer-based trajectory prediction network for end-to-end autonomous driving without rules called Target-point Attention Transformer network (TAT). We use the attention mechanism to realize the interaction between the predicted trajectory and the perception features as well as target-points. We demonstrate that our proposed method outperforms existing conditional imitation learning and GRU-based methods, significantly reducing the occurrence of accidents and improving route completion. We evaluate our approach in complex closed loop driving scenarios in cities using the CARLA simulator and achieve state-of-the-art performance.

cs.RO↗

arXiv2023

3DRP-Net: 3D Relative Position-aware Network for 3D Visual Grounding

3D visual grounding aims to localize the target object in a 3D point cloud by a free-form language description. Typically, the sentences describing the target object tend to provide information about its relative relation between other objects and its position within the whole scene. In this work, we propose a relation-aware one-stage framework, named 3D Relative Position-aware Network (3DRP-Net), which can effectively capture the relative spatial relationships between objects and enhance object attributes. Specifically, 1) we propose a 3D Relative Position Multi-head Attention (3DRP-MA) module to analyze relative relations from different directions in the context of object pairs, which helps the model to focus on the specific object relations mentioned in the sentence. 2) We designed a soft-labeling strategy to alleviate the spatial ambiguity caused by redundant points, which further stabilizes and enhances the learning process through a constant and discriminative distribution. Extensive experiments conducted on three benchmarks (i.e., ScanRefer and Nr3D/Sr3D) demonstrate that our method outperforms all the state-of-the-art methods in general. The source code will be released on GitHub.

cs.CV↗

arXiv2023

Non-volatile Phase-only Transmissive Spatial Light Modulators

Free-space modulation of light is crucial for many applications, from light detection and ranging to virtual or augmented reality. Traditional means of modulating free-space light involves spatial light modulators based on liquid crystals and microelectromechanical systems, which are bulky, have large pixel areas (~10 micron x 10 micron), and require high driving voltage. Recent progress in meta-optics has shown promise to circumvent some of the limitations. By integrating active materials with sub-wavelength pixels in a meta-optic, the power consumption can be dramatically reduced while achieving a faster speed. However, these reconfiguration methods are volatile and hence require constant application of control signals, leading to phase jitter and crosstalk. Additionally, to control a large number of pixels, it is essential to implement a memory within each pixel to have a tractable number of control signals. Here, we develop a device with nonvolatile, electrically programmable, phase-only modulation of free-space infrared radiation in transmission using the low-loss phase-change material (PCM) Sb2Se3. By coupling an ultra-thin PCM layer to a high quality (Q)-factor (Q~406) diatomic metasurface, we demonstrate a phase-only modulation of ~0.25pi (~0.2pi) in simulation (experiment), ten times larger than a bare PCM layer of the same thickness. The device shows excellent endurance over 1,000 switching cycles. We then advance the device geometry, to enable independent control of 17 meta-molecules, achieving ten deterministic resonance levels with a 2pi phase shift. By independently controlling the phase delay of pixels, we further show tunable far-field beam shaping. Our work paves the way to realizing non-volatile transmissive phase-only spatial light modulators.

physics.optics↗

arXiv2023

Photon-assisted Landau Zener transitions in a tunable driven Rabi dimer coupled to a micromechanical resonator

Employing the multiple Davydov D$_2$ Ansatz with the time-dependent variational principle, we have investigated photon-assisted Landau-Zener (LZ) transitions and qubit manipulation in a hybrid quantum electrodynamics device. Modelled as a Rabi dimer, the device comprises of two interacting transmission-line resonators, each coupled to a qubit. The qubits, driven by independent harmonic fields, are further modulated by a micromechanical resonator mimicked by a phonon mode. The impacts of two independent driving fields on the qubit dynamics are carefully examined. The energy diagram of the system and the photon number mobilization on the resonators are analyzed to explain the behaviour of the LZ transitions and qubit dynamics while taking into account the influence of the single phonon mode. Results show that low phonon frequencies can alter the qubit dynamics, particularly in the absence of the driving fields, {and a strong phonon coupling strength can significantly perturb the qubit dynamics thanks to a high influx of phonon energy}. Notably, only the photon frequency affects the oscillation frequency of qubit polarization. This study unveils the imperative roles that photons and phonons play in the Rabi dimer model.

quant-ph↗

arXiv2023

Towards Authentic Face Restoration with Iterative Diffusion Models and Beyond

An authentic face restoration system is becoming increasingly demanding in many computer vision applications, e.g., image enhancement, video communication, and taking portrait. Most of the advanced face restoration models can recover high-quality faces from low-quality ones but usually fail to faithfully generate realistic and high-frequency details that are favored by users. To achieve authentic restoration, we propose $\textbf{IDM}$, an $\textbf{I}$teratively learned face restoration system based on denoising $\textbf{D}$iffusion $\textbf{M}$odels (DDMs). We define the criterion of an authentic face restoration system, and argue that denoising diffusion models are naturally endowed with this property from two aspects: intrinsic iterative refinement and extrinsic iterative enhancement. Intrinsic learning can preserve the content well and gradually refine the high-quality details, while extrinsic enhancement helps clean the data and improve the restoration task one step further. We demonstrate superior performance on blind face restoration tasks. Beyond restoration, we find the authentically cleaned data by the proposed restoration system is also helpful to image generation tasks in terms of training stabilization and sample quality. Without modifying the models, we achieve better quality than state-of-the-art on FFHQ and ImageNet generation using either GANs or diffusion models.

cs.CV↗