Searcharxiv⌕ Search

arXiv subjects

Jinjing Zhu

Publications and source records attributed to Jinjing Zhu.

At least 19 recordsLinked to original sources
arXiv2026

Sat2City v2: Native 3D City Asset Generation from a Single Satellite Image

Generating explicit textured 3D city assets from a single satellite image is important for urban simulation and digital twins. Most prior methods, however, learn 3D proxies optimized for rendering street-view images or videos over prescribed viewpoints and trajectories, rather than producing explicit 3D assets. Our previous framework, Sat2City, took a first step toward this goal with task-specific cascaded sparse-voxel latent diffusion conditioned on satellite-derived height maps. However, it relied on synthetic data and did not condition appearance on the satellite input. To address these limitations, we present Sat2City v2, a framework for generating explicit textured 3D city assets directly from real satellite images. We first construct a real-world dataset of 16,027 geographically matched but weakly aligned satellite-image-textured-mesh pairs, comprising 14,651 training pairs and 1,376 held-out test pairs. Building on this dataset, Sat2City v2 uses a native structured-latent 3D prior pretrained on curated assets to anchor geospatial adaptation to its asset manifold. We empirically demonstrate that weakly aligned satellite images and regional textured 3D meshes can support asset-level conditional generation: learned attention lets satellite tokens steer geometry generation without calibrated pixel-to-surface correspondence, while the generated geometry anchors satellite-guided material synthesis. Sat2City v2 ranks first among the evaluated baselines on every reported metric across geometry accuracy, generative mesh quality, and satellite-to-asset feature alignment.

cs.CV↗
arXiv2026

CR-Refiner: An Object-Centric Optimal Transport Reranker for Edit-Conditioned 3D Scene Retrieval

Edit-conditioned 3D scene retrieval pairs a reference 3D room with a natural-language modification and retrieves rooms from a corpus that satisfy the edit. Three lines of prior work each fall short on this task. 2D composed image retrieval reasons over pixel-level edits and has no primitive for 3D object sets. 3D foundation encoders embed individual objects but cannot compose at the scene level. 3D scene-grounding methods localize references inside a static scene rather than rank modified rooms across a corpus. We present CR-Refiner, a training-free reranker that wraps any base retriever's top-K candidates with three components. A frozen LLM parses the edit into a structured query entity, and each candidate is scored by an unbalanced optimal-transport problem over a 1xG cost matrix coupling category, style, material, and geometry. The unbalanced solver lets the single-entity query drop mass on irrelevant objects, modelling the asymmetry directly. An axis-conditional structural prior adds size-keyword cues for geometric edits and subject-anchor direction cues for spatial edits. An LLM verifier refines the top three candidates with continuous confidence. Because no benchmark evaluates compositional matching over 3D object sets, we additionally release 3D-CER, 4,963 edit-conditioned queries over a 23,381-room indoor corpus across five edit axes, with multi-positive ground truth, CIRR-style hard subsets, and zero-target adversarials. Across three qualitatively distinct base retrievers, CR-Refiner consistently improves hard-subset R@1 and mAP@10 on every edit axis.

cs.CV↗
arXiv2026

EventVGGT: Exploring Cross-Modal Distillation for Consistent Event-based Depth Estimation

Event cameras offer superior sensitivity to high-speed motion and extreme lighting, making event-based monocular depth estimation a promising approach for robust 3D perception in challenging conditions. However, progress is severely hindered by the scarcity of dense depth annotations. While recent annotation-free approaches mitigate this by distilling knowledge from Vision Foundation Models (VFMs), a critical limitation persists: they process event streams as independent frames. By neglecting the inherent temporal continuity of event data, these methods fail to leverage the rich temporal priors encoded in VFMs, ultimately yielding temporally inconsistent and less accurate depth predictions. To address this, we introduce EventVGGT, a novel framework that explicitly models the event stream as a coherent video sequence. To the best of our knowledge, we are the first to distill spatio-temporal and multi-view geometric priors from the Visual Geometry Grounded Transformer (VGGT) into the event domain. We achieve this via a comprehensive tri-level distillation strategy: (i) Cross-Modal Feature Mixture (CMFM) bridges the modality gap at the output level by fusing RGB and event features to generate auxiliary depth predictions; (ii) Spatio-Temporal Feature Distillation (STFD) distills VGGT's powerful spatio-temporal representations at the feature level; and (iii) Temporal Consistency Distillation (TCD) enforces cross-frame coherence at the temporal level by aligning inter-frame depth changes. Extensive experiments demonstrate that EventVGGT consistently outperforms existing methods -- reducing the absolute mean depth error at 30m by over 53\% on EventScape (from 2.30 to 1.06) -- while exhibiting robust zero-shot generalization on the unseen DENSE and MVSEC datasets. The code is available at https://github.com/yinruiRen/EventVGGT.

cs.CV↗
arXiv2026

Seeing Through Multiple Views: Parameter-Efficient Fine-Tuning via Selective Neurons for Consistent Radiology Report Generation

Recent years have seen substantial advances in radiology report generation (RRG), yet existing approaches predominantly adopt direct feature fusion when handling multi-view X-ray images. Such approaches overlook the potential clinical inconsistencies and inaccuracies arising when a single model processes different views, adversely impacting performance and clinical reliability. To this end, we introduce View-PNDF (View-specific Pattern Neuron Detection and Fine-tuning), a parameter-efficient framework that fosters view-consistent report generation from a neuronal perspective. Specifically, View-PNDF comprises: (i) a view-specific neuron detection module identifying neurons responsive to particular views, (ii) a verification module quantifying the existence of these neurons, and (iii) a selective fine-tuning strategy strengthening detected neurons while preserving view-agnostic representations. By updating only view-specific neurons, View-PNDF achieves consistent diagnoses across different views with reduced computational costs. Subsequently, we employ Large Language Models (LLMs) to consolidate the view-specific reports into a complete radiology report. Furthermore, we use traditional Natural Language Generation (NLG) metrics-based assessment on integrated reports for baseline comparison and employ LLM-based assessment (e.g., GPT-4o) on view-specific reports to capture clinical significance. Extensive experiments on two medical RRG benchmarks demonstrate that View-PNDF substantially improves view-specific chest X-ray report generation quality while maintaining robust general-view performance.

cs.CV↗
arXiv2026

Bridging the Post-discharge Gap: A Traceable Multi-agent Framework for Safe and Continuous Care

Post-discharge clinical follow-up is critical for maintaining continuity of care and mitigating long-term health risks. However, traditional follow-up paradigms suffer from shortage of health workforce, fragmented patient histories, and information silos across clinical departments. While large language models have demonstrated potential in medical question-answering, their deployment in continuous care is hindered by hallucination risks and a fundamental inability to reason over longitudinal, patient-specific constraints. Here we present Healink, a memory-enhanced multi-agent framework to support AI-assisted post-discharge follow-up by generating prescription-grounded, traceable responses that improved completeness and perceived clinical utility in retrospective and physician-blinded evaluations. The architecture seamlessly integrates a triage routing mechanism, a unified memory enhancement module utilizing a robust relational database for optimal latency, and a strict constraint-based retrieval-augmented generation engine. By vectorizing historical clinical records and employing weighted similarity functions across diverse phenotypic and intervention dimensions, Healink ensures precise inter-patient and intra-patient case matching while actively preventing cross-departmental drug conflicts. We evaluated Healink on a dataset comprising 400 continuous and 85 highly complex real-world follow-up cases, alongside the webMedQA benchmark. In a rigorous single-blind evaluation conducted by clinical experts, the framework outperformed human physician baselines in both authoritativeness and clinical safety. By generating a traceable, white-box evidence chain, Healink provides a scalable, safe, and highly effective paradigm for intelligent patient management, ultimately enhancing societal healthcare outcomes.

cs.MA↗
arXiv2026

Beyond a Single Light: A Large-Scale Aerial Dataset for Urban Scene Reconstruction Under Varying Illumination

Recent advances in Neural Radiance Fields and 3D Gaussian Splatting have demonstrated strong potential for large-scale UAV-based 3D reconstruction tasks by fitting the appearance of images. However, real-world large-scale captures are often based on multi-temporal data capture, where illumination inconsistencies across different times of day can significantly lead to color artifacts, geometric inaccuracies, and inconsistent appearance. Due to the lack of UAV datasets that systematically capture the same areas under varying illumination conditions, this challenge remains largely underexplored. To fill this gap, we introduceSkyLume, a large-scale, real-world UAV dataset specifically designed for studying illumination robust 3D reconstruction in urban scene modeling: (1) We collect data from 10 urban regions data comprising more than 100k high resolution UAV images (four oblique views and nadir), where each region is captured at three periods of the day to systematically isolate illumination changes. (2) To support precise evaluation of geometry and appearance, we provide per-scene LiDAR scans and accurate 3D ground-truth for assessing depth, surface normals, and reconstruction quality under varying illumination. (3) For the inverse rendering task, we introduce the Temporal Consistency Coefficient (TCC), a metric that measuress cross-time albedo stability and directly evaluates the robustness of the disentanglement of light and material. We aim for this resource to serve as a foundation that advances research and real-world evaluation in large-scale inverse rendering, geometry reconstruction, and novel view synthesis.

cs.CV↗
arXiv2026

Holo360D: A Large-Scale Real-World Dataset with Continuous Trajectories for Advancing Panoramic 3D Reconstruction and Beyond

While feed-forward 3D reconstruction models have advanced rapidly, they still exhibit degraded performance on panoramas due to spherical distortions. Moreover, existing panoramic 3D datasets are predominantly collected with 360 cameras fixed at discrete locations, resulting in discontinuous trajectories. These limitations critically hinder the development of panoramic feed-forward 3D reconstruction, especially for the multi-view setting. In this paper, we present Holo360D, a comprehensive dataset containing 109,495 panoramas paired with registered point clouds, meshes, and aligned camera poses. To our knowledge, Holo360D is the first large-scale dataset that provides continuous panoramic sequences with accurately aligned high-completeness depth maps. The raw data are initially collected using a 3D laser scanner coupled with a 360 camera. Subsequently, the raw data are processed with both online and offline SLAM systems. Furthermore, to enhance the 3D data quality, a post-processing pipeline tailored for the 360 dataset is proposed, including geometry denoising, mesh hole filling, and region-specific remeshing. Finally, we establish a new benchmark by fine-tuning 3D reconstruction models on Holo360D, providing key insights into effective fine-tuning strategies. Our results demonstrate that Holo360D delivers superior training signals and provides a comprehensive benchmark for advancing panoramic 3D reconstruction models. Datasets and Code will be made publicly available.

cs.CV↗
arXiv2026

SRTJ: Self-Evolving Rule-Driven Training-Free LLM Jailbreaking

LLMs are increasingly equipped with safety alignment mechanisms, yet recent studies demonstrate that they remain vulnerable to jailbreaking attacks that elicit harmful behaviors without explicit policy violations. While a growing body of work has explored automated jailbreak strategies, existing methods face several fundamental challenges, including the lack of systematic utilization of both successful and failed attack experiences, as well as the absence of principled mechanisms for composing and selecting reusable attack rules under diverse constraints. As a result, existing methods struggle to accumulate transferable knowledge over time and to reliably adapt attack strategies across different targets and evolving safety mechanisms. To address these issues, we propose a Self-Evolving Rule-Driven Training-Free Jailbreak (SRTJ) framework that systematically discovers, composes, and refines attack strategies through interaction and feedback, without updating model parameters. Specifically, SRTJ couples experience-driven attack generation with answer set programming (ASP)-based rule selection and constraint-aware composition, where iterative verifier feedback is leveraged to jointly refine successful strategies and analyze failure patterns. The resulting rule memory evolves in a hierarchical multi-level manner, explicitly organizing distilled attack knowledge into long-term, middle-term, and short-term rules, thereby capturing both stable transferable strategies and transient adaptive behaviors to effectively balance exploration and exploitation across attack attempts. Extensive experiments on mainstream jailbreak benchmark (HarmBench) demonstrate that SRTJ achieves strong and stable attack performance across different target LLMs, while exhibiting improved robustness and generalization compared to existing jailbreak methods. The code is available at https://github.com/TheSolkatt/SRTJ.

cs.CR↗
arXiv2026

Exploring the Vulnerabilities of Federated Learning: A Deep Dive into Gradient Inversion Attacks

Federated Learning (FL) has emerged as a promising privacy-preserving collaborative model training paradigm without sharing raw data. However, recent studies have revealed that private information can still be leaked through shared gradient information and attacked by Gradient Inversion Attacks (GIA). While many GIA methods have been proposed, a detailed analysis, evaluation, and summary of these methods are still lacking. Although various survey papers summarize existing privacy attacks in FL, few studies have conducted extensive experiments to unveil the effectiveness of GIA and their associated limiting factors in this context. To fill this gap, we first undertake a systematic review of GIA and categorize existing methods into three types, i.e., \textit{optimization-based} GIA (OP-GIA), \textit{generation-based} GIA (GEN-GIA), and \textit{analytics-based} GIA (ANA-GIA). Then, we comprehensively analyze and evaluate the three types of GIA in FL, providing insights into the factors that influence their performance, practicality, and potential threats. Our findings indicate that OP-GIA is the most practical attack setting despite its unsatisfactory performance, while GEN-GIA has many dependencies and ANA-GIA is easily detectable, making them both impractical. Finally, we offer a three-stage defense pipeline to users when designing FL frameworks and protocols for better privacy protection and share some future research directions from the perspectives of attackers and defenders that we believe should be pursued. We hope that our study can help researchers design more robust FL frameworks to defend against these attacks.

cs.CR↗
arXiv2025

PanDA: Towards Panoramic Depth Anything with Unlabeled Panoramas and Mobius Spatial Augmentation

Recently, Depth Anything Models (DAMs) - a type of depth foundation models - have demonstrated impressive zero-shot capabilities across diverse perspective images. Despite its success, it remains an open question regarding DAMs' performance on panorama images that enjoy a large field-of-view (180x360) but suffer from spherical distortions. To address this gap, we conduct an empirical analysis to evaluate the performance of DAMs on panoramic images and identify their limitations. For this, we undertake comprehensive experiments to assess the performance of DAMs from three key factors: panoramic representations, 360 camera positions for capturing scenarios, and spherical spatial transformations. This way, we reveal some key findings, e.g., DAMs are sensitive to spatial transformations. We then propose a semi-supervised learning (SSL) framework to learn a panoramic DAM, dubbed PanDA. Under the umbrella of SSL, PanDA first learns a teacher model by fine-tuning DAM through joint training on synthetic indoor and outdoor panoramic datasets. Then, a student model is trained using large-scale unlabeled data, leveraging pseudo-labels generated by the teacher model. To enhance PanDA's generalization capability, M"obius transformation-based spatial augmentation (MTSA) is proposed to impose consistency regularization between the predicted depth maps from the original and spatially transformed ones. This subtly improves the student model's robustness to various spatial transformations, even under severe distortions. Extensive experiments demonstrate that PanDA exhibits remarkable zero-shot capability across diverse scenes, and outperforms the data-specific panoramic depth estimation methods on two popular real-world benchmarks.

cs.CV↗
arXiv2025

Rethinking Knowledge in Distillation: An In-context Sample Retrieval Perspective

Conventional knowledge distillation (KD) approaches are designed for the student model to predict similar output as the teacher model for each sample. Unfortunately, the relationship across samples with same class is often neglected. In this paper, we explore to redefine the knowledge in distillation, capturing the relationship between each sample and its corresponding in-context samples (a group of similar samples with the same or different classes), and perform KD from an in-context sample retrieval perspective. As KD is a type of learned label smoothing regularization (LSR), we first conduct a theoretical analysis showing that the teacher's knowledge from the in-context samples is a crucial contributor to regularize the student training with the corresponding samples. Buttressed by the analysis, we propose a novel in-context knowledge distillation (IC-KD) framework that shows its superiority across diverse KD paradigms (offline, online, and teacher-free KD). Firstly, we construct a feature memory bank from the teacher model and retrieve in-context samples for each corresponding sample through retrieval-based learning. We then introduce Positive In-Context Distillation (PICD) to reduce the discrepancy between a sample from the student and the aggregated in-context samples with the same class from the teacher in the logit space. Moreover, Negative In-Context Distillation (NICD) is introduced to separate a sample from the student and the in-context samples with different classes from the teacher in the logit space. Extensive experiments demonstrate that IC-KD is effective across various types of KD, and consistently achieves state-of-the-art performance on CIFAR-100 and ImageNet datasets.

cs.CV↗
arXiv2024

CLIP the Divergence: Language-guided Unsupervised Domain Adaptation

Unsupervised domain adaption (UDA) has emerged as a popular solution to tackle the divergence between the labeled source and unlabeled target domains. Recently, some research efforts have been made to leverage large vision-language models, such as CLIP, and then fine-tune or learn prompts from them for addressing the challenging UDA task. In this work, we shift the gear to a new direction by directly leveraging CLIP to measure the domain divergence and propose a novel language-guided approach for UDA, dubbed as CLIP-Div. Our key idea is to harness CLIP to 1) measure the domain divergence via the acquired domain-agnostic distribution and 2) calibrate the target pseudo labels with language guidance, to effectively reduce the domain gap and improve the UDA model's generalization capability. Specifically, our major technical contribution lies in the proposed two novel language-guided domain divergence measurement losses: absolute divergence and relative divergence. These loss terms furnish precise guidelines for aligning the distributions of the source and target domains with the domain-agnostic distribution derived from CLIP. Additionally, we propose a language-guided pseudo-labeling strategy for calibrating the target pseudo labels. Buttressed by it, we show that a further implementation for self-training can enhance the UDA model's generalization capability on the target domain. CLIP-Div surpasses state-of-the-art CNN-based methods by a substantial margin, achieving a performance boost of +10.3% on Office-Home, +1.5% on Office-31, +0.2% on VisDA-2017, and +24.3% on DomainNet, respectively.

cs.CV↗
arXiv2024

Towards Dynamic and Small Objects Refinement for Unsupervised Domain Adaptative Nighttime Semantic Segmentation

Nighttime semantic segmentation plays a crucial role in practical applications, such as autonomous driving, where it frequently encounters difficulties caused by inadequate illumination conditions and the absence of well-annotated datasets. Moreover, semantic segmentation models trained on daytime datasets often face difficulties in generalizing effectively to nighttime conditions. Unsupervised domain adaptation (UDA) has shown the potential to address the challenges and achieved remarkable results for nighttime semantic segmentation. However, existing methods still face limitations in 1) their reliance on style transfer or relighting models, which struggle to generalize to complex nighttime environments, and 2) their ignorance of dynamic and small objects like vehicles and poles, which are difficult to be directly learned from other domains. This paper proposes a novel UDA method that refines both label and feature levels for dynamic and small objects for nighttime semantic segmentation. First, we propose a dynamic and small object refinement module to complement the knowledge of dynamic and small objects from the source domain to target the nighttime domain. These dynamic and small objects are normally context-inconsistent in under-exposed conditions. Then, we design a feature prototype alignment module to reduce the domain gap by deploying contrastive learning between features and prototypes of the same class from different domains, while re-weighting the categories of dynamic and small objects. Extensive experiments on three benchmark datasets demonstrate that our method outperforms prior arts by a large margin for nighttime segmentation. Project page: https://rorisis.github.io/DSRNSS/.

cs.CV↗
arXiv2024

Energy-based Domain-Adaptive Segmentation with Depth Guidance

Recent endeavors have been made to leverage self-supervised depth estimation as guidance in unsupervised domain adaptation (UDA) for semantic segmentation. Prior arts, however, overlook the discrepancy between semantic and depth features, as well as the reliability of feature fusion, thus leading to suboptimal segmentation performance. To address this issue, we propose a novel UDA framework called SMART (croSs doMain semAntic segmentation based on eneRgy esTimation) that utilizes Energy-Based Models (EBMs) to obtain task-adaptive features and achieve reliable feature fusion for semantic segmentation with self-supervised depth estimates. Our framework incorporates two novel components: energy-based feature fusion (EB2F) and energy-based reliable fusion Assessment (RFA) modules. The EB2F module produces task-adaptive semantic and depth features by explicitly measuring and reducing their discrepancy using Hopfield energy for better feature fusion. The RFA module evaluates the reliability of the feature fusion using an energy score to improve the effectiveness of depth guidance. Extensive experiments on two datasets demonstrate that our method achieves significant performance gains over prior works, validating the effectiveness of our energy-based learning approach.

cs.CV↗
arXiv2024

Source-Free Cross-Modal Knowledge Transfer by Unleashing the Potential of Task-Irrelevant Data

Source-free cross-modal knowledge transfer is a crucial yet challenging task, which aims to transfer knowledge from one source modality (e.g., RGB) to the target modality (e.g., depth or infrared) with no access to the task-relevant (TR) source data due to memory and privacy concerns. A recent attempt leverages the paired task-irrelevant (TI) data and directly matches the features from them to eliminate the modality gap. However, it ignores a pivotal clue that the paired TI data could be utilized to effectively estimate the source data distribution and better facilitate knowledge transfer to the target modality. To this end, we propose a novel yet concise framework to unlock the potential of paired TI data for enhancing source-free cross-modal knowledge transfer. Our work is buttressed by two key technical components. Firstly, to better estimate the source data distribution, we introduce a Task-irrelevant data-Guided Modality Bridging (TGMB) module. It translates the target modality data (e.g., infrared) into the source-like RGB images based on paired TI data and the guidance of the available source model to alleviate two key gaps: 1) inter-modality gap between the paired TI data; 2) intra-modality gap between TI and TR target data. We then propose a Task-irrelevant data-Guided Knowledge Transfer (TGKT) module that transfers knowledge from the source model to the target model by leveraging the paired TI data. Notably, due to the unavailability of labels for the TR target data and its less reliable prediction from the source model, our TGKT model incorporates a self-supervised pseudo-labeling approach to enable the target model to learn from its predictions. Extensive experiments show that our method achieves state-of-the-art performance on three datasets (RGB-to-depth and RGB-to-infrared).

cs.CV↗
arXiv2023

A Unified Framework for Unsupervised Domain Adaptation based on Instance Weighting

Despite the progress made in domain adaptation, solving Unsupervised Domain Adaptation (UDA) problems with a general method under complex conditions caused by label shifts between domains remains a formidable task. In this work, we comprehensively investigate four distinct UDA settings including closed set domain adaptation, partial domain adaptation, open set domain adaptation, and universal domain adaptation, where shared common classes between source and target domains coexist alongside domain-specific private classes. The prominent challenges inherent in diverse UDA settings center around the discrimination of common/private classes and the precise measurement of domain discrepancy. To surmount these challenges effectively, we propose a novel yet effective method called Learning Instance Weighting for Unsupervised Domain Adaptation (LIWUDA), which caters to various UDA settings. Specifically, the proposed LIWUDA method constructs a weight network to assign weights to each instance based on its probability of belonging to common classes, and designs Weighted Optimal Transport (WOT) for domain alignment by leveraging instance weights. Additionally, the proposed LIWUDA method devises a Separate and Align (SA) loss to separate instances with low similarities and align instances with high similarities. To guide the learning of the weight network, Intra-domain Optimal Transport (IOT) is proposed to enforce the weights of instances in common classes to follow a uniform distribution. Through the integration of those three components, the proposed LIWUDA method demonstrates its capability to address all four UDA settings in a unified manner. Experimental evaluations conducted on three benchmark datasets substantiate the effectiveness of the proposed LIWUDA method.

cs.CV↗
arXiv2023

Test-Time Adaptation for Nighttime Color-Thermal Semantic Segmentation

The ability to scene understanding in adverse visual conditions, e.g., nighttime, has sparked active research for RGB-Thermal (RGB-T) semantic segmentation. However, it is essentially hampered by two critical problems: 1) the day-night gap of RGB images is larger than that of thermal images, and 2) the class-wise performance of RGB images at night is not consistently higher or lower than that of thermal images. we propose the first test-time adaptation (TTA) framework, dubbed Night-TTA, to address the problems for nighttime RGBT semantic segmentation without access to the source (daytime) data during adaptation. Our method enjoys three key technical parts. Firstly, as one modality (e.g., RGB) suffers from a larger domain gap than that of the other (e.g., thermal), Imaging Heterogeneity Refinement (IHR) employs an interaction branch on the basis of RGB and thermal branches to prevent cross-modal discrepancy and performance degradation. Then, Class Aware Refinement (CAR) is introduced to obtain reliable ensemble logits based on pixel-level distribution aggregation of the three branches. In addition, we also design a specific learning scheme for our TTA framework, which enables the ensemble logits and three student logits to collaboratively learn to improve the quality of predictions during the testing phase of our Night TTA. Extensive experiments show that our method achieves state-of-the-art (SoTA) performance with a 13.07% boost in mIoU.

cs.CV↗
arXiv2023

A Good Student is Cooperative and Reliable: CNN-Transformer Collaborative Learning for Semantic Segmentation

In this paper, we strive to answer the question "how to collaboratively learn convolutional neural network (CNN)-based and vision transformer (ViT)-based models by selecting and exchanging the reliable knowledge between them for semantic segmentation?" Accordingly, we propose an online knowledge distillation (KD) framework that can simultaneously learn compact yet effective CNN-based and ViT-based models with two key technical breakthroughs to take full advantage of CNNs and ViT while compensating their limitations. Firstly, we propose heterogeneous feature distillation (HFD) to improve students' consistency in low-layer feature space by mimicking heterogeneous features between CNNs and ViT. Secondly, to facilitate the two students to learn reliable knowledge from each other, we propose bidirectional selective distillation (BSD) that can dynamically transfer selective knowledge. This is achieved by 1) region-wise BSD determining the directions of knowledge transferred between the corresponding regions in the feature space and 2) pixel-wise BSD discerning which of the prediction knowledge to be transferred in the logit space. Extensive experiments on three benchmark datasets demonstrate that our proposed framework outperforms the state-of-the-art online distillation methods by a large margin, and shows its efficacy in learning collaboratively between ViT-based and CNN-based models.

cs.CV↗