arXiv subjects

Humphrey Shi

Publications and source records attributed to Humphrey Shi.

At least 37 records · Page 2Linked to original sources

arXiv2025

T2I-Copilot: A Training-Free Multi-Agent Text-to-Image System for Enhanced Prompt Interpretation and Interactive Generation

Text-to-Image (T2I) generative models have revolutionized content creation but remain highly sensitive to prompt phrasing, often requiring users to repeatedly refine prompts multiple times without clear feedback. While techniques such as automatic prompt engineering, controlled text embeddings, denoising, and multi-turn generation mitigate these issues, they offer limited controllability, or often necessitate additional training, restricting the generalization abilities. Thus, we introduce T2I-Copilot, a training-free multi-agent system that leverages collaboration between (Multimodal) Large Language Models to automate prompt phrasing, model selection, and iterative refinement. This approach significantly simplifies prompt engineering while enhancing generation quality and text-image alignment compared to direct generation. Specifically, T2I-Copilot consists of three agents: (1) Input Interpreter, which parses the input prompt, resolves ambiguities, and generates a standardized report; (2) Generation Engine, which selects the appropriate model from different types of T2I models and organizes visual and textual prompts to initiate generation; and (3) Quality Evaluator, which assesses aesthetic quality and text-image alignment, providing scores and feedback for potential regeneration. T2I-Copilot can operate fully autonomously while also supporting human-in-the-loop intervention for fine-grained control. On GenAI-Bench, using open-source generation models, T2I-Copilot achieves a VQA score comparable to commercial models RecraftV3 and Imagen 3, surpasses FLUX1.1-pro by 6.17% at only 16.59% of its cost, and outperforms FLUX.1-dev and SD 3.5 Large by 9.11% and 6.36%. Code will be released at: https://github.com/SHI-Labs/T2I-Copilot.

cs.CV↗

arXiv2025

Efficient Image Generation with Variadic Attention Heads

While the integration of transformers in vision models have yielded significant improvements on vision tasks they still require significant amounts of computation for both training and inference. Restricted attention mechanisms significantly reduce these computational burdens but come at the cost of losing either global or local coherence. We propose a simple, yet powerful method to reduce these trade-offs: allow the attention heads of a single transformer to attend to multiple receptive fields. We demonstrate our method utilizing Neighborhood Attention (NA) and integrate it into a StyleGAN based architecture for image generation. With this work, dubbed StyleNAT, we are able to achieve a FID of 2.05 on FFHQ, a 6% improvement over StyleGAN-XL, while utilizing 28% fewer parameters and with 4$\times$ the throughput capacity. StyleNAT achieves the Pareto Frontier on FFHQ-256 and demonstrates powerful and efficient image generation on other datasets. Our code and model checkpoints are publicly available at: https://github.com/SHI-Labs/StyleNAT

cs.CV↗

arXiv2025

Distilling Normalizing Flows

Explicit density learners are becoming an increasingly popular technique for generative models because of their ability to better model probability distributions. They have advantages over Generative Adversarial Networks due to their ability to perform density estimation and having exact latent-variable inference. This has many advantages, including: being able to simply interpolate, calculate sample likelihood, and analyze the probability distribution. The downside of these models is that they are often more difficult to train and have lower sampling quality. Normalizing flows are explicit density models, that use composable bijective functions to turn an intractable probability function into a tractable one. In this work, we present novel knowledge distillation techniques to increase sampling quality and density estimation of smaller student normalizing flows. We seek to study the capacity of knowledge distillation in Compositional Normalizing Flows to understand the benefits and weaknesses provided by these architectures. Normalizing flows have unique properties that allow for a non-traditional forms of knowledge transfer, where we can transfer that knowledge within intermediate layers. We find that through this distillation, we can make students significantly smaller while making substantial performance gains over a non-distilled student. With smaller models there is a proportionally increased throughput as this is dependent upon the number of bijectors, and thus parameters, in the network.

cs.LG↗

arXiv2025

Person Recognition at Altitude and Range: Fusion of Face, Body Shape and Gait

We address the problem of whole-body person recognition in unconstrained environments. This problem arises in surveillance scenarios such as those in the IARPA Biometric Recognition and Identification at Altitude and Range (BRIAR) program, where biometric data is captured at long standoff distances, elevated viewing angles, and under adverse atmospheric conditions (e.g., turbulence and high wind velocity). To this end, we propose FarSight, a unified end-to-end system for person recognition that integrates complementary biometric cues across face, gait, and body shape modalities. FarSight incorporates novel algorithms across four core modules: multi-subject detection and tracking, recognition-aware video restoration, modality-specific biometric feature encoding, and quality-guided multi-modal fusion. These components are designed to work cohesively under degraded image conditions, large pose and scale variations, and cross-domain gaps. Extensive experiments on the BRIAR dataset, one of the most comprehensive benchmarks for long-range, multi-modal biometric recognition, demonstrate the effectiveness of FarSight. Compared to our preliminary system, this system achieves a 34.1% absolute gain in 1:1 verification accuracy (TAR@0.1% FAR), a 17.8% increase in closed-set identification (Rank-20), and a 34.3% reduction in open-set identification errors (FNIR@1% FPIR). Furthermore, FarSight was evaluated in the 2025 NIST RTE Face in Video Evaluation (FIVE), which conducts standardized face recognition testing on the BRIAR dataset. These results establish FarSight as a state-of-the-art solution for operational biometric recognition in challenging real-world conditions.

cs.CV↗

arXiv2025

Generalized Neighborhood Attention: Multi-dimensional Sparse Attention at the Speed of Light

Many sparse attention mechanisms such as Neighborhood Attention have typically failed to consistently deliver speedup over the self attention baseline. This is largely due to the level of complexity in attention infrastructure, and the rapid evolution of AI hardware architecture. At the same time, many state-of-the-art foundational models, particularly in computer vision, are heavily bound by attention, and need reliable sparsity to escape the O(n^2) complexity. In this paper, we study a class of promising sparse attention mechanisms that focus on locality, and aim to develop a better analytical model of their performance improvements. We first introduce Generalized Neighborhood Attention (GNA), which can describe sliding window, strided sliding window, and blocked attention. We then consider possible design choices in implementing these approaches, and create a simulator that can provide much more realistic speedup upper bounds for any given setting. Finally, we implement GNA on top of a state-of-the-art fused multi-headed attention (FMHA) kernel designed for the NVIDIA Blackwell architecture in CUTLASS. Our implementation can fully realize the maximum speedup theoretically possible in many perfectly block-sparse cases, and achieves an effective utilization of 1.3 petaFLOPs/second in FP16. In addition, we plug various GNA configurations into off-the-shelf generative models, such as Cosmos-7B, HunyuanVideo, and FLUX, and show that it can deliver 28% to 46% end-to-end speedup on B200 without any fine-tuning. We will open source our simulator and Blackwell kernels directly through the NATTEN project.

cs.CV↗

arXiv2025

StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text

Text-to-video diffusion models enable the generation of high-quality videos that follow text instructions, making it easy to create diverse and individual content. However, existing approaches mostly focus on high-quality short video generation (typically 16 or 24 frames), ending up with hard-cuts when naively extended to the case of long video synthesis. To overcome these limitations, we introduce StreamingT2V, an autoregressive approach for long video generation of 80, 240, 600, 1200 or more frames with smooth transitions. The key components are:(i) a short-term memory block called conditional attention module (CAM), which conditions the current generation on the features extracted from the previous chunk via an attentional mechanism, leading to consistent chunk transitions, (ii) a long-term memory block called appearance preservation module, which extracts high-level scene and object features from the first video chunk to prevent the model from forgetting the initial scene, and (iii) a randomized blending approach that enables to apply a video enhancer autoregressively for infinitely long videos without inconsistencies between chunks. Experiments show that StreamingT2V generates high motion amount. In contrast, all competing image-to-video methods are prone to video stagnation when applied naively in an autoregressive manner. Thus, we propose with StreamingT2V a high-quality seamless text-to-long video generator that outperforms competitors with consistency and motion. Our code will be available at: https://github.com/Picsart-AI-Research/StreamingT2V

cs.CV↗

arXiv2025

RAGME: Retrieval Augmented Video Generation for Enhanced Motion Realism

Video generation is experiencing rapid growth, driven by advances in diffusion models and the development of better and larger datasets. However, producing high-quality videos remains challenging due to the high-dimensional data and the complexity of the task. Recent efforts have primarily focused on enhancing visual quality and addressing temporal inconsistencies, such as flickering. Despite progress in these areas, the generated videos often fall short in terms of motion complexity and physical plausibility, with many outputs either appearing static or exhibiting unrealistic motion. In this work, we propose a framework to improve the realism of motion in generated videos, exploring a complementary direction to much of the existing literature. Specifically, we advocate for the incorporation of a retrieval mechanism during the generation phase. The retrieved videos act as grounding signals, providing the model with demonstrations of how the objects move. Our pipeline is designed to apply to any text-to-video diffusion model, conditioning a pretrained model on the retrieved samples with minimal fine-tuning. We demonstrate the superiority of our approach through established metrics, recently proposed benchmarks, and qualitative results, and we highlight additional applications of the framework.

cs.CV↗

arXiv2025

Slow-Fast Architecture for Video Multi-Modal Large Language Models

Balancing temporal resolution and spatial detail under limited compute budget remains a key challenge for video-based multi-modal large language models (MLLMs). Existing methods typically compress video representations using predefined rules before feeding them into the LLM, resulting in irreversible information loss and often ignoring input instructions. To address this, we propose a novel slow-fast architecture that naturally circumvents this trade-off, enabling the use of more input frames while preserving spatial details. Inspired by how humans first skim a video before focusing on relevant parts, our slow-fast design employs a dual-token strategy: 1) "fast" visual tokens -- a compact set of compressed video features -- are fed into the LLM alongside text embeddings to provide a quick overview; 2) "slow" visual tokens -- uncompressed video features -- are cross-attended by text embeddings through specially designed hybrid decoder layers, enabling instruction-aware extraction of relevant visual details with linear complexity. We conduct systematic exploration to optimize both the overall architecture and key components. Experiments show that our model significantly outperforms self-attention-only baselines, extending the input capacity from 16 to 128 frames with just a 3% increase in computation, and achieving a 16% average performance improvement across five video understanding benchmarks. Our 7B model achieves state-of-the-art performance among models of similar size. Furthermore, our slow-fast architecture is a plug-and-play design that can be integrated into other video MLLMs to improve efficiency and scalability.

cs.CV↗

arXiv2025

ClassDiffusion: More Aligned Personalization Tuning with Explicit Class Guidance

Recent text-to-image customization works have proven successful in generating images of given concepts by fine-tuning diffusion models on a few examples. However, tuning-based methods inherently tend to overfit the concepts, resulting in failure to create the concept under multiple conditions (*e.g.*, headphone is missing when generating "a `dog wearing a headphone"). Interestingly, we notice that the base model before fine-tuning exhibits the capability to compose the base concept with other elements (*e.g.*, "a dog wearing a headphone"), implying that the compositional ability only disappears after personalization tuning. We observe a semantic shift in the customized concept after fine-tuning, indicating that the personalized concept is not aligned with the original concept, and further show through theoretical analyses that this semantic shift leads to increased difficulty in sampling the joint conditional probability distribution, resulting in the loss of the compositional ability. Inspired by this finding, we present **ClassDiffusion**, a technique that leverages a **semantic preservation loss** to explicitly regulate the concept space when learning a new concept. Although simple, this approach effectively prevents semantic drift during the fine-tuning process of the target concepts. Extensive qualitative and quantitative experiments demonstrate that the use of semantic preservation loss effectively improves the compositional abilities of fine-tuning models. Lastly, we also extend our ClassDiffusion to personalized video generation, demonstrating its flexibility.

cs.CV↗

arXiv2025

Eagle: Exploring The Design Space for Multimodal LLMs with Mixture of Encoders

The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks.

cs.CV↗

arXiv2024

Combating Label Noise With A General Surrogate Model For Sample Selection

Modern deep learning systems are data-hungry. Learning with web data is one of the feasible solutions, but will introduce label noise inevitably, which can hinder the performance of deep neural networks. Sample selection is an effective way to deal with label noise. The key is to separate clean samples based on some criterion. Previous methods pay more attention to the small loss criterion where small-loss samples are regarded as clean ones. Nevertheless, such a strategy relies on the learning dynamics of each data instance. Some noisy samples are still memorized due to frequently occurring corrupted learning patterns. To tackle this problem, a training-free surrogate model is preferred, freeing from the effect of memorization. In this work, we propose to leverage the vision-language surrogate model CLIP to filter noisy samples automatically. CLIP brings external knowledge to facilitate the selection of clean samples with its ability of text-image alignment. Furthermore, a margin adaptive loss is designed to regularize the selection bias introduced by CLIP, providing robustness to label noise. We validate the effectiveness of our proposed method on both real-world and synthetic noisy datasets. Our method achieves significant improvement without CLIP involved during the inference stage.

cs.CV↗

arXiv2024

CompactFlowNet: Efficient Real-time Optical Flow Estimation on Mobile Devices

We present CompactFlowNet, the first real-time mobile neural network for optical flow prediction, which involves determining the displacement of each pixel in an initial frame relative to the corresponding pixel in a subsequent frame. Optical flow serves as a fundamental building block for various video-related tasks, such as video restoration, motion estimation, video stabilization, object tracking, action recognition, and video generation. While current state-of-the-art methods prioritize accuracy, they often overlook constraints regarding speed and memory usage. Existing light models typically focus on reducing size but still exhibit high latency, compromise significantly on quality, or are optimized for high-performance GPUs, resulting in sub-optimal performance on mobile devices. This study aims to develop a mobile-optimized optical flow model by proposing a novel mobile device-compatible architecture, as well as enhancements to the training pipeline, which optimize the model for reduced weight, low memory utilization, and increased speed while maintaining minimal error. Our approach demonstrates superior or comparable performance to the state-of-the-art lightweight models on the challenging KITTI and Sintel benchmarks. Furthermore, it attains a significantly accelerated inference speed, thereby yielding real-time operational efficiency on the iPhone 8, while surpassing real-time performance levels on more advanced mobile devices.

cs.CV↗

arXiv2024

Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment

Test-time adaptation (TTA) aims to improve the performance of source-domain pre-trained models on previously unseen, shifted target domains. Traditional TTA methods primarily adapt model weights based on target data streams, making model performance sensitive to the amount and order of target data. The recently proposed diffusion-driven TTA methods mitigate this by adapting model inputs instead of weights, where an unconditional diffusion model, trained on the source domain, transforms target-domain data into a synthetic domain that is expected to approximate the source domain. However, in this paper, we reveal that although the synthetic data in diffusion-driven TTA seems indistinguishable from the source data, it is unaligned with, or even markedly different from the latter for deep networks. To address this issue, we propose a \textbf{S}ynthetic-\textbf{D}omain \textbf{A}lignment (SDA) framework. Our key insight is to fine-tune the source model with synthetic data to ensure better alignment. Specifically, we first employ a conditional diffusion model to generate labeled samples, creating a synthetic dataset. Subsequently, we use the aforementioned unconditional diffusion model to add noise to and denoise each sample before fine-tuning. This Mix of Diffusion (MoD) process mitigates the potential domain misalignment between the conditional and unconditional models. Extensive experiments across classifiers, segmenters, and multimodal large language models (MLLMs, \eg, LLaVA) demonstrate that SDA achieves superior domain alignment and consistently outperforms existing diffusion-driven TTA methods. Our code is available at https://github.com/SHI-Labs/Diffusion-Driven-Test-Time-Adaptation-via-Synthetic-Domain-Alignment.

cs.CV↗

arXiv2024

Diffusion for Natural Image Matting

We aim to leverage diffusion to address the challenging image matting task. However, the presence of high computational overhead and the inconsistency of noise sampling between the training and inference processes pose significant obstacles to achieving this goal. In this paper, we present DiffMatte, a solution designed to effectively overcome these challenges. First, DiffMatte decouples the decoder from the intricately coupled matting network design, involving only one lightweight decoder in the iterations of the diffusion process. With such a strategy, DiffMatte mitigates the growth of computational overhead as the number of samples increases. Second, we employ a self-aligned training strategy with uniform time intervals, ensuring a consistent noise sampling between training and inference across the entire time domain. Our DiffMatte is designed with flexibility in mind and can seamlessly integrate into various modern matting architectures. Extensive experimental results demonstrate that DiffMatte not only reaches the state-of-the-art level on the Composition-1k test set, surpassing the best methods in the past by 5% and 15% in the SAD metric and MSE metric respectively, but also show stronger generalization ability in other benchmarks.

cs.CV↗

arXiv2024

Learning Trimaps via Clicks for Image Matting

Despite significant advancements in image matting, existing models heavily depend on manually-drawn trimaps for accurate results in natural image scenarios. However, the process of obtaining trimaps is time-consuming, lacking user-friendliness and device compatibility. This reliance greatly limits the practical application of all trimap-based matting methods. To address this issue, we introduce Click2Trimap, an interactive model capable of predicting high-quality trimaps and alpha mattes with minimal user click inputs. Through analyzing real users' behavioral logic and characteristics of trimaps, we successfully propose a powerful iterative three-class training strategy and a dedicated simulation function, making Click2Trimap exhibit versatility across various scenarios. Quantitative and qualitative assessments on synthetic and real-world matting datasets demonstrate Click2Trimap's superior performance compared to all existing trimap-free matting methods. Especially, in the user study, Click2Trimap achieves high-quality trimap and matting predictions in just an average of 5 seconds per image, demonstrating its substantial practical value in real-world applications.

cs.CV↗

arXiv2024

Collaborative Vision-Text Representation Optimizing for Open-Vocabulary Segmentation

Pre-trained vision-language models, e.g. CLIP, have been increasingly used to address the challenging Open-Vocabulary Segmentation (OVS) task, benefiting from their well-aligned vision-text embedding space. Typical solutions involve either freezing CLIP during training to unilaterally maintain its zero-shot capability, or fine-tuning CLIP vision encoder to achieve perceptual sensitivity to local regions. However, few of them incorporate vision-text collaborative optimization. Based on this, we propose the Content-Dependent Transfer to adaptively enhance each text embedding by interacting with the input image, which presents a parameter-efficient way to optimize the text representation. Besides, we additionally introduce a Representation Compensation strategy, reviewing the original CLIP-V representation as compensation to maintain the zero-shot capability of CLIP. In this way, the vision and text representation of CLIP are optimized collaboratively, enhancing the alignment of the vision-text feature space. To the best of our knowledge, we are the first to establish the collaborative vision-text optimizing mechanism within the OVS field. Extensive experiments demonstrate our method achieves superior performance on popular OVS benchmarks. In open-vocabulary semantic segmentation, our method outperforms the previous state-of-the-art approaches by +0.5, +2.3, +3.4, +0.4 and +1.1 mIoU, respectively on A-847, A-150, PC-459, PC-59 and PAS-20. Furthermore, in a panoptic setting on ADE20K, we achieve the performance of 27.1 PQ, 73.5 SQ, and 32.9 RQ. Code will be available at https://github.com/jiaosiyu1999/MAFT-Plus.git .

cs.CV↗

arXiv2024

Faster Neighborhood Attention: Reducing the O(n^2) Cost of Self Attention at the Threadblock Level

Neighborhood attention reduces the cost of self attention by restricting each token's attention span to its nearest neighbors. This restriction, parameterized by a window size and dilation factor, draws a spectrum of possible attention patterns between linear projection and self attention. Neighborhood attention, and more generally sliding window attention patterns, have long been bounded by infrastructure, particularly in higher-rank spaces (2-D and 3-D), calling for the development of custom kernels, which have been limited in either functionality, or performance, if not both. In this work, we aim to massively improve upon existing infrastructure by providing two new methods for implementing neighborhood attention. We first show that neighborhood attention can be represented as a batched GEMM problem, similar to standard attention, and implement it for 1-D and 2-D neighborhood attention. These kernels on average provide 895% and 272% improvement in full precision runtime compared to existing naive CUDA kernels for 1-D and 2-D neighborhood attention respectively. We find that aside from being heavily bound by memory bandwidth, certain inherent inefficiencies exist in all unfused implementations of neighborhood attention, which in most cases undo their theoretical efficiency gain. Motivated by the progress made into fused dot-product attention kernels, we developed fused neighborhood attention; an adaptation of fused dot-product attention kernels that allow fine-grained control over attention across different spatial axes. Known for reducing the quadratic time complexity of self attention to a linear complexity, neighborhood attention can now enjoy a reduced and constant memory footprint, and record-breaking half precision runtime. We observe that our fused implementation successfully circumvents some of the unavoidable inefficiencies in unfused implementations...

cs.CV↗

arXiv2024

GradBias: Unveiling Word Influence on Bias in Text-to-Image Generative Models

Recent progress in Text-to-Image (T2I) generative models has enabled high-quality image generation. As performance and accessibility increase, these models are gaining significant attraction and popularity: ensuring their fairness and safety is a priority to prevent the dissemination and perpetuation of biases. However, existing studies in bias detection focus on closed sets of predefined biases (e.g., gender, ethnicity). In this paper, we propose a general framework to identify, quantify, and explain biases in an open set setting, i.e. without requiring a predefined set. This pipeline leverages a Large Language Model (LLM) to propose biases starting from a set of captions. Next, these captions are used by the target generative model for generating a set of images. Finally, Vision Question Answering (VQA) is leveraged for bias evaluation. We show two variations of this framework: OpenBias and GradBias. OpenBias detects and quantifies biases, while GradBias determines the contribution of individual prompt words on biases. OpenBias effectively detects both well-known and novel biases related to people, objects, and animals and highly aligns with existing closed-set bias detection methods and human judgment. GradBias shows that neutral words can significantly influence biases and it outperforms several baselines, including state-of-the-art foundation models. Code available here: https://github.com/Moreno98/GradBias.

cs.CV↗