arXiv subjects

Jinjia Zhou

Publications and source records attributed to Jinjia Zhou.

At least 19 recordsLinked to original sources

arXiv2026

Flux as a Codec: Zero-Shot Lossy Compression with Flux

We propose FluxC, a zero-shot codec that uses Flux as a compression prior. Although DiffC has shown that pretrained diffusion models can serve as zero-shot codecs, we ask how the choice of coding path and reconstruction method affects rate--distortion--perception behavior. We analyze FluxC along three axes: coding cost and bit allocation, reconstruction method, and low rate distortion. At matched signal-to-noise ratio (SNR), we show that the path choice does not affect the ideal information rate, but changes how bits are allocated across time steps. For reconstruction, we compare ancestral sampling, one-step ODE reconstruction, and many-step ODE reconstruction. Our analysis shows that ODE-based reconstruction can achieve lower distortion than ancestral sampling, especially in the low rate regime. Experiments on Gaussian sources and natural images support this prediction. Moreover, FluxC performs competitively with, and in some cases outperforms, recent neural image compression methods, suggesting that Flux can serve as a principled zero-shot codec.

cs.LG↗

arXiv2026

PASTA: Prompt-Aware Annealed Diffusion Sampling

While Diffusion models produce high-quality images, real-world visual design requires intentionally preserving clean regions for headlines, logos, or UI elements. Existing layout-aware text-to-image methods struggle to strictly preserve reserved regions, often introducing unwanted content that complicates downstream editing. Existing approaches either manipulate attention to displace content or suppress conditioning via noise-level interventions. However, these methods sacrifice visual coherence, producing flat regions or noticeable style inconsistencies between reserved and non-reserved areas. To address this limitation, we propose a Prompt-Aware Shared-Latent Training-Free Annealed Diffusion Sampling (PASTA), a training-free framework that enforces reserved-region constraints by selectively applying semantic guidance to object-allowed regions during sampling. PASTA operates on a shared latent trajectory, where full-prompt and context-prompt guidance are fused through mask-guided sampling. An annealed schedule suppresses semantic leakage during early denoising stages while gradually restoring visual consistency in later stages. Extensive experiments demonstrate that PASTA improves reserved-region fidelity while preserving overall image realism and perceptual consistency.

cs.CV↗

arXiv2026

Where Did the Light Come From? Diagnosing Cast-Shadow Reasoning in VLMs and Borrowing the Missing Shadow Prior from Image Generation Models

Shadows are observable consequences of latent physical causes, arising from the interaction between illumination, object geometry, and scene structure. A model that understands shadows should be able not only to read a shadow's direction, but also to infer where a shadow should fall when it is absent, and to judge whether a shadow is physically consistent with the scene. We introduce ShadowBench, built from real photographs, which poses these three questions to vision-language models (VLMs). From each scene, we build a correct-shadow, a shadow-removed, and a human-annotated wrong-direction state. Evaluating 19 open VLMs (2B to 78B) and frontier proprietary models, we find that VLMs read a visible shadow's direction well above chance and improve with scale, yet infer the absent direction only weakly and fall to chance at judging physical consistency even under a bias-free two-alternative test that removes the yes-bias of single-image naturalness questions. Only Gemini-3.1-pro exceeds chance on validity, and even then only modestly and far below humans. Chain-of-Thought (CoT) does not help. The missing ingredient is not reasoning but knowledge that image generation models (IGMs) already hold. Routing a VLM through an IGM lifts its inference well above its direct attempt and helps it flag the wrong-direction shadows it otherwise accepts. Cast-shadow direction is a single facet of illumination understanding, but the sharp gap it reveals between VLMs and IGMs suggests image-generation or reconstruction objectives as a route toward illumination-aware VLMs.

cs.CV↗

arXiv2026

Astrolabe: Spherical-Map Guidance Across Diffusion Pipelines for Full-Body Capture from Unconstrained Images

Full-body capture from unconstrained photographs requires global correspondence across arbitrary views, poses, crops, and occlusions. Yet pose, geometry, and foundation features estimated in this setting are too unreliable for dense matching or appearance transfer, while diffusion rectifiers and optimization pipelines expose no common interface for consuming such uncertain correspondence. Our insight is that correspondence need not be locally accurate: its coarse viewpoint and body layout can still organize how a diffusion prior adapts and guides reconstruction. We introduce \emph{Astrolabe}, a host-portable adapter built on frozen viewpoint-guided spherical maps (SPH). A fixed bounded transform converts SPH into a spatial noise shift, which is matched during prior adaptation and reused during downstream denoising or score-distillation guidance in both pipeline categories. When a rectifier exposes a reference router, the same target/reference SPH additionally supplies coarse compatibility scores to select native appearance features; router-free optimization uses only the shared shift path. Astrolabe therefore follows one SPH--shift--adapt--guide process without dense warping or a learned control branch. Across Puzzle-IOI and 4D-Dress, it improves all reported image metrics in both hosts and all paired Puzzle-IOI geometry metrics; image gains extend to rear views, while 4D-Dress geometry remains stable overall.

cs.CV↗

arXiv2026

Few-step Generative Models as Lossy Compression

DiffC provides a principled way to reuse pre-trained diffusion models for lossy compression, but its encoding and decoding procedures remain slow because they require many discretized forward and reverse steps. We study whether few-step generative models -- Rectified Flow, Consistency Trajectory Models (CTM), and MeanFlow -- can be cast as codecs within the same reverse channel coding (RCC) framework. The main challenge is that RCC requires posterior and shared distribution parameters, whereas these models do not explicitly parameterize intermediate conditional distributions. For Rectified Flow and MeanFlow, we use the equivalence between velocity parameterization and diffusion-style denoising parameterization to derive the quantities required by RCC. For CTM, which is distilled from EDM, we adopt the EDM noise parameterization together with local Gaussian approximations of the sender and shared distributions at intermediate states. This yields a proof-of-concept probabilistic formulation that enables compression with pre-trained few-step generative models without retraining. On low-resolution benchmarks, the resulting codecs reduce encoding and decoding time and improve realism in the low-bit-rate regime.

cs.CV↗

arXiv2026

Coarse Semantic Injection for LLM-Conditioned Structured Indoor Prediction

Large language models (LLMs) have recently been used as structured decoders for indoor understanding from 3D point-token inputs. However, point cloud encoders often under-represent thin structural elements such as doors and windows after voxelization and sparse pooling, and may miss individual furniture instances in cluttered scenes. We propose an interface-preserving semantic augmentation for LLM-conditioned structured decoding. The key idea is to associate semantic evidence with the point-cloud representation, reduce it to a coarse four-group code (furniture, walls, openings, and others), and encode it as an RGBB point interface: red for furniture, green for walls, blue for openings, and black for others, where RGBB denotes four semantic color states represented in three RGB channels rather than an additional fourth channel. This semantic color code is appended to the original raw point attributes before tokenization, so geometry and semantics share the same sparse tokenization path while the downstream language model decoder and output serialization remain unchanged. We further introduce a lightweight routed semantic shift module, with an auxiliary head used only for training-time ratio/budget regularization and analysis, to strengthen semantic cues after sparse pooling. The overall pipeline can use RGB-derived semantic evidence. Under these controlled semantic-source settings, the reported metrics improve across Structured3D, the SpatialLM dataset, and ARKitScenes, especially for opening localization and per-instance furniture detection in cluttered scenes. Ablations clarify the roles of semantic source, color coding, token fusion, and shift injection, while also showing that color/entropy effects remain nontrivial.

cs.CV↗

arXiv2025

GRACE: Designing Generative Face Video Codec via Agile Hardware-Centric Workflow

The Animation-based Generative Codec (AGC) is an emerging paradigm for talking-face video compression. However, deploying its intricate decoder on resource and power-constrained edge devices presents challenges due to numerous parameters, the inflexibility to adapt to dynamically evolving algorithms, and the high power consumption induced by extensive computations and data transmission. This paper for the first time proposes a novel field programmable gate arrays (FPGAs)-oriented AGC deployment scheme for edge-computing video services. Initially, we analyze the AGC algorithm and employ network compression methods including post-training static quantization and layer fusion techniques. Subsequently, we design an overlapped accelerator utilizing the co-processor paradigm to perform computations through software-hardware co-design. The hardware processing unit comprises engines such as convolution, grid sampling, upsample, etc. Parallelization optimization strategies like double-buffered pipelines and loop unrolling are employed to fully exploit the resources of FPGA. Ultimately, we establish an AGC FPGA prototype on the PYNQ-Z1 platform using the proposed scheme, achieving \textbf{24.9$\times$} and \textbf{4.1$\times$} higher energy efficiency against commercial Central Processing Unit (CPU) and Graphic Processing Unit (GPU), respectively. Specifically, only \textbf{11.7} microjoules ($\upmu$J) are required for one pixel reconstructed by this FPGA system.

cs.CV↗

arXiv2025

Adaptive Sampling Scheduler

Consistent distillation methods have evolved into effective techniques that significantly accelerate the sampling process of diffusion models. Although existing methods have achieved remarkable results, the selection of target timesteps during distillation mainly relies on deterministic or stochastic strategies, which often require sampling schedulers to be designed specifically for different distillation processes. Moreover, this pattern severely limits flexibility, thereby restricting the full sampling potential of diffusion models in practical applications. To overcome these limitations, this paper proposes an adaptive sampling scheduler that is applicable to various consistency distillation frameworks. The scheduler introduces three innovative strategies: (i) dynamic target timestep selection, which adapts to different consistency distillation frameworks by selecting timesteps based on their computed importance; (ii) Optimized alternating sampling along the solution trajectory by guiding forward denoising and backward noise addition based on the proposed time step importance, enabling more effective exploration of the solution space to enhance generation performance; and (iii) Utilization of smoothing clipping and color balancing techniques to achieve stable and high-quality generation results at high guidance scales, thereby expanding the applicability of consistency distillation models in complex generation scenarios. We validated the effectiveness and flexibility of the adaptive sampling scheduler across various consistency distillation methods through comprehensive experimental evaluations. Experimental results consistently demonstrated significant improvements in generative performance, highlighting the strong adaptability achieved by our method.

cs.CV↗

arXiv2025

TopKD: Top-scaled Knowledge Distillation

Recent advances in knowledge distillation (KD) predominantly emphasize feature-level knowledge transfer, frequently overlooking critical information embedded within the teacher's logit distributions. In this paper, we revisit logit-based distillation and reveal an underexplored yet critical element: Top-K knowledge. Motivated by this insight, we propose Top-scaled Knowledge Distillation (TopKD), a simple, efficient, and architecture-agnostic framework that significantly enhances logit-based distillation. TopKD consists of two main components: (1) a Top-K Scaling Module (TSM), which adaptively amplifies the most informative logits, and (2) a Top-K Decoupled Loss (TDL), which offers targeted and effective supervision. Notably, TopKD integrates seamlessly into existing KD methods without introducing extra modules or requiring architectural changes. Extensive experiments on CIFAR-100, ImageNet, STL-10, and Tiny-ImageNet demonstrate that TopKD consistently surpasses state-of-the-art distillation methods. Moreover, our method demonstrates substantial effectiveness when distilling Vision Transformers, underscoring its versatility across diverse network architectures. These findings highlight the significant potential of logits to advance knowledge distillation.

cs.CV↗

arXiv2025

TKG-DM: Training-free Chroma Key Content Generation Diffusion Model

Diffusion models have enabled the generation of high-quality images with a strong focus on realism and textual fidelity. Yet, large-scale text-to-image models, such as Stable Diffusion, struggle to generate images where foreground objects are placed over a chroma key background, limiting their ability to separate foreground and background elements without fine-tuning. To address this limitation, we present a novel Training-Free Chroma Key Content Generation Diffusion Model (TKG-DM), which optimizes the initial random noise to produce images with foreground objects on a specifiable color background. Our proposed method is the first to explore the manipulation of the color aspects in initial noise for controlled background generation, enabling precise separation of foreground and background without fine-tuning. Extensive experiments demonstrate that our training-free method outperforms existing methods in both qualitative and quantitative evaluations, matching or surpassing fine-tuned models. Finally, we successfully extend it to other tasks (e.g., consistency models and text-to-video), highlighting its transformative potential across various generative applications where independent control of foreground and background is crucial.

cs.CV↗

arXiv2025

Audio-Visual Driven Compression for Low-Bitrate Talking Head Videos

Talking head video compression has advanced with neural rendering and keypoint-based methods, but challenges remain, especially at low bit rates, including handling large head movements, suboptimal lip synchronization, and distorted facial reconstructions. To address these problems, we propose a novel audio-visual driven video codec that integrates compact 3D motion features and audio signals. This approach robustly models significant head rotations and aligns lip movements with speech, improving both compression efficiency and reconstruction quality. Experiments on the CelebV-HQ dataset show that our method reduces bitrate by 22% compared to VVC and by 8.5% over state-of-the-art learning-based codec. Furthermore, it provides superior lip-sync accuracy and visual fidelity at comparable bitrates, highlighting its effectiveness in bandwidth-constrained scenarios.

eess.IV↗

arXiv2025

Bidirectional Learned Facial Animation Codec for Low Bitrate Talking Head Videos

Existing deep facial animation coding techniques efficiently compress talking head videos by applying deep generative models. Instead of compressing the entire video sequence, these methods focus on compressing only the keyframe and the keypoints of non-keyframes (target frames). The target frames are then reconstructed by utilizing a single keyframe, and the keypoints of the target frame. Although these unidirectional methods can reduce the bitrate, they rely on a single keyframe and often struggle to capture large head movements accurately, resulting in distortions in the facial region. In this paper, we propose a novel bidirectional learned animation codec that generates natural facial videos using past and future keyframes. First, in the Bidirectional Reference-Guided Auxiliary Stream Enhancement (BRG-ASE) process, we introduce a compact auxiliary stream for non-keyframes, which is enhanced by adaptively selecting one of two keyframes (past and future). This stream improves video quality with a slight increase in bitrate. Then, in the Bidirectional Reference-Guided Video Reconstruction (BRG-VRec) process, we animate the adaptively selected keyframe and reconstruct the target frame using both the animated keyframe and the auxiliary frame. Extensive experiments demonstrate a 55% bitrate reduction compared to the latest animation based video codec, and a 35% bitrate reduction compared to the latest video coding standard, Versatile Video Coding (VVC) on a talking head video dataset. It showcases the efficiency of our approach in improving video quality while simultaneously decreasing bitrate.

eess.IV↗

arXiv2025

GenAIReading: Augmenting Human Cognition with Interactive Digital Textbooks Using Large Language Models and Image Generation Models

Cognitive augmentation is a cornerstone in advancing education, particularly through personalized learning. However, personalizing extensive textual materials, such as narratives and academic textbooks, remains challenging due to their heavy use, which can hinder learner engagement and understanding. Building on cognitive theories like Dual Coding Theory -- which posits that combining textual and visual information enhances comprehension and memory -- this study explores the potential of Generative AI (GenAI) to enrich educational materials. We utilized large language models (LLMs) to generate concise text summaries and image generation models (IGMs) to create visually aligned content from textual inputs. After recruiting 24 participants, we verified that integrating AI-generated supplementary materials significantly improved learning outcomes, increasing post-reading test scores by 7.50%. These findings underscore GenAI's transformative potential in creating adaptive learning environments that enhance cognitive augmentation.

cs.HC↗

arXiv2025

Multi-perspective Contrastive Logit Distillation

In previous studies on knowledge distillation, the significance of logit distillation has frequently been overlooked. To revitalize logit distillation, we present a novel perspective by reconsidering its computation based on the semantic properties of logits and exploring how to utilize it more efficiently. Logits often contain a substantial amount of high-level semantic information; however, the conventional approach of employing logits to compute Kullback-Leibler (KL) divergence does not account for their semantic properties. Furthermore, this direct KL divergence computation fails to fully exploit the potential of logits. To address these challenges, we introduce a novel and efficient logit distillation method, Multi-perspective Contrastive Logit Distillation (MCLD), which substantially improves the performance and efficacy of logit distillation. In comparison to existing logit distillation methods and complex feature distillation methods, MCLD attains state-of-the-art performance in image classification, and transfer learning tasks across multiple datasets, including CIFAR-100, ImageNet, Tiny-ImageNet, and STL-10. Additionally, MCLD exhibits superior training efficiency and outstanding performance with distilling on Vision Transformers, further emphasizing its notable advantages. This study unveils the vast potential of logits in knowledge distillation and seeks to offer valuable insights for future research.

cs.CV↗

arXiv2024

Unicorn: Unified Neural Image Compression with One Number Reconstruction

Prevalent lossy image compression schemes can be divided into: 1) explicit image compression (EIC), including traditional standards and neural end-to-end algorithms; 2) implicit image compression (IIC) based on implicit neural representations (INR). The former is encountering impasses of either leveling off bitrate reduction at a cost of tremendous complexity while the latter suffers from excessive smoothing quality as well as lengthy decoder models. In this paper, we propose an innovative paradigm, which we dub \textbf{Unicorn} (\textbf{U}nified \textbf{N}eural \textbf{I}mage \textbf{C}ompression with \textbf{O}ne \textbf{N}number \textbf{R}econstruction). By conceptualizing the images as index-image pairs and learning the inherent distribution of pairs in a subtle neural network model, Unicorn can reconstruct a visually pleasing image from a randomly generated noise with only one index number. The neural model serves as the unified decoder of images while the noises and indexes corresponds to explicit representations. As a proof of concept, we propose an effective and efficient prototype of Unicorn based on latent diffusion models with tailored model designs. Quantitive and qualitative experimental results demonstrate that our prototype achieves significant bitrates reduction compared with EIC and IIC algorithms. More impressively, benefitting from the unified decoder, our compression ratio escalates as the quantity of images increases. We envision that more advanced model designs will endow Unicorn with greater potential in image compression. We will release our codes in \url{https://github.com/uniqzheng/Unicorn-Laduree}.

cs.CV↗

arXiv2024

Visual question answering based evaluation metrics for text-to-image generation

Text-to-image generation and text-guided image manipulation have received considerable attention in the field of image generation tasks. However, the mainstream evaluation methods for these tasks have difficulty in evaluating whether all the information from the input text is accurately reflected in the generated images, and they mainly focus on evaluating the overall alignment between the input text and the generated images. This paper proposes new evaluation metrics that assess the alignment between input text and generated images for every individual object. Firstly, according to the input text, chatGPT is utilized to produce questions for the generated images. After that, we use Visual Question Answering(VQA) to measure the relevance of the generated images to the input text, which allows for a more detailed evaluation of the alignment compared to existing methods. In addition, we use Non-Reference Image Quality Assessment(NR-IQA) to evaluate not only the text-image alignment but also the quality of the generated images. Experimental results show that our proposed evaluation approach is the superior metric that can simultaneously assess finer text-image alignment and image quality while allowing for the adjustment of these ratios.

cs.CV↗

arXiv2024

Block based Adaptive Compressive Sensing with Sampling Rate Control

Compressive sensing (CS), acquiring and reconstructing signals below the Nyquist rate, has great potential in image and video acquisition to exploit data redundancy and greatly reduce the amount of sampled data. To further reduce the sampled data while keeping the video quality, this paper explores the temporal redundancy in video CS and proposes a block based adaptive compressive sensing framework with a sampling rate (SR) control strategy. To avoid redundant compression of non-moving regions, we first incorporate moving block detection between consecutive frames, and only transmit the measurements of moving blocks. The non-moving regions are reconstructed from the previous frame. In addition, we propose a block storage system and a dynamic threshold to achieve adaptive SR allocation to each frame based on the area of moving regions and target SR for controlling the average SR within the target SR. Finally, to reduce blocking artifacts and improve reconstruction quality, we adopt a cooperative reconstruction of the moving and non-moving blocks by referring to the measurements of the non-moving blocks from the previous frame. Extensive experiments have demonstrated that this work is able to control SR and obtain better performance than existing works.

cs.CV↗

arXiv2024

Learn from Balance: Rectifying Knowledge Transfer for Long-Tailed Scenarios

Knowledge Distillation (KD) transfers knowledge from a large pre-trained teacher network to a compact and efficient student network, making it suitable for deployment on resource-limited media terminals. However, traditional KD methods require balanced data to ensure robust training, which is often unavailable in practical applications. In such scenarios, a few head categories occupy a substantial proportion of examples. This imbalance biases the trained teacher network towards the head categories, resulting in severe performance degradation on the less represented tail categories for both the teacher and student networks. In this paper, we propose a novel framework called Knowledge Rectification Distillation (KRDistill) to address the imbalanced knowledge inherited in the teacher network through the incorporation of the balanced category priors. Furthermore, we rectify the biased predictions produced by the teacher network, particularly focusing on the tail categories. Consequently, the teacher network can provide balanced and accurate knowledge to train a reliable student network. Intensive experiments conducted on various long-tailed datasets demonstrate that our KRDistill can effectively train reliable student networks in realistic scenarios of data imbalance.

cs.CV↗