Di Xu

Lingdong Kong, Yaru Niu, Shaoyuan Xie et al.

Accurate depth estimation under out-of-distribution (OoD) scenarios, such as adverse weather conditions, sensor failure, and noise contamination, is desirable for safety-critical applications. Existing depth estimation systems, however, suffer inevitably from real-world corruptions and perturbations and are struggled to provide reliable depth predictions under such cases. In this paper, we summarize the winning solutions from the RoboDepth Challenge -- an academic competition designed to facilitate and advance robust OoD depth estimation. This challenge was developed based on the newly established KITTI-C and NYUDepth2-C benchmarks. We hosted two stand-alone tracks, with an emphasis on robust self-supervised and robust fully-supervised depth estimation, respectively. Out of more than two hundred participants, nine unique and top-performing solutions have appeared, with novel designs ranging from the following aspects: spatial- and frequency-domain augmentations, masked image modeling, image restoration and super-resolution, adversarial training, diffusion-based noise suppression, vision-language pre-training, learned model ensembling, and hierarchical feature enhancement. Extensive experimental analyses along with insightful observations are drawn to better understand the rationale behind each design. We hope this challenge could lay a solid foundation for future research on robust and reliable depth estimation and beyond. The datasets, competition toolkit, workshop recordings, and source code from the winning teams are publicly available on the challenge website.

Haoran Wang, Di Xu, Dongliang He et al.

Video-text retrieval (VTR) is an attractive yet challenging task for multi-modal understanding, which aims to search for relevant video (text) given a query (video). Existing methods typically employ completely heterogeneous visual-textual information to align video and text, whilst lacking the awareness of homogeneous high-level semantic information residing in both modalities. To fill this gap, in this work, we propose a novel visual-linguistic aligning model named HiSE for VTR, which improves the cross-modal representation by incorporating explicit high-level semantics. First, we explore the hierarchical property of explicit high-level semantics, and further decompose it into two levels, i.e. discrete semantics and holistic semantics. Specifically, for visual branch, we exploit an off-the-shelf semantic entity predictor to generate discrete high-level semantics. In parallel, a trained video captioning model is employed to output holistic high-level semantics. As for the textual modality, we parse the text into three parts including occurrence, action and entity. In particular, the occurrence corresponds to the holistic high-level semantics, meanwhile both action and entity represent the discrete ones. Then, different graph reasoning techniques are utilized to promote the interaction between holistic and discrete high-level semantics. Extensive experiments demonstrate that, with the aid of explicit high-level semantics, our method achieves the superior performance over state-of-the-art methods on three benchmark datasets, including MSR-VTT, MSVD and DiDeMo.

Yinghui Xing, Xiaoting Su, Shizhou Zhang et al.

Infrared imaging plays a critical role in low-light and adverse weather conditions. However, due to the distinct characteristics of infrared images, existing foundation models such as Masked Autoencoder (MAE) trained on visible data perform suboptimal in infrared image interpretation tasks. To bridge this gap, an infrared foundation model known as InfMAE was developed and pre-trained on large-scale infrared datasets. Despite its effectiveness, InfMAE still faces several limitations, including the omission of informative tokens, insufficient modeling of global associations, and neglect of non-uniform noise. In this paper, we propose a Dual-domain Guided Infrared foundation model based on MAE (DuGI-MAE). First, we design a deterministic masking strategy based on token entropy, preserving only high-entropy tokens for reconstruction to enhance informativeness. Next, we introduce a Dual-Domain Guidance (DDG) module, which simultaneously captures global token relationships and adaptively filters non-uniform background noise commonly present in infrared imagery. To facilitate large-scale pretraining, we construct Inf-590K, a comprehensive infrared image dataset encompassing diverse scenes, various target types, and multiple spatial resolutions. Pretrained on Inf-590K, DuGI-MAE demonstrates strong generalization capabilities across various downstream tasks, including infrared object detection, semantic segmentation, and small target detection. Experimental results validate the superiority of the proposed method over both supervised and self-supervised comparison methods. Our code is available in the supplementary material.

Di Xu, Qifan Xu, Kevin Nhieu et al.

Suppression of thoracic bone shadows on chest X-rays (CXRs) has been indicated to improve the diagnosis of pulmonary disease. Previous approaches can be categorized as unsupervised physical and supervised deep learning models. Nevertheless, with physical models able to preserve morphological details but at the cost of extremely long processing time, existing DL methods face challenges of gathering sufficient/qualitative ground truth (GT) for robust training, thus leading to failure in maintaining clinically acceptable false positive rates. We hereby propose a generalizable yet efficient workflow of two stages: (1) training pairs generation with GT bone shadows eliminated in by a physical model in spatially transformed gradient fields. (2) fully supervised image denoising network training on stage-one datasets for fast rib removal on incoming CXRs. For step two, we designed a densely connected network called SADXNet, combined with peak signal to noise ratio and multi-scale structure similarity index measure objective minimization to suppress bony structures. The SADXNet organizes spatial filters in U shape (e.g., X=7; filters = 16, 64, 256, 512, 256, 64, 16) and preserves the feature map dimension throughout the network flow. Visually, SADXNet can suppress the rib edge and that near the lung wall/vertebra without jeopardizing the vessel/abnormality conspicuity. Quantitively, it achieves RMSE of ~0 during testing with one prediction taking <1s. Downstream tasks including lung nodule detection as well as common lung disease classification and localization are used to evaluate our proposed rib suppression mechanism. We observed 3.23% and 6.62% area under the curve (AUC) increase as well as 203 and 385 absolute false positive decrease for lung nodule detection and common lung disease localization, separately.

Michael J. Ryan, Yanzhe Zhang, Amol Salunkhe et al. · gatech

Evaluating user-facing AI applications remains a central challenge, especially in open-ended domains such as travel planning, clinical note generation, or dialogue. The gold standard is user feedback (e.g., thumbs up/down) or behavioral signals (e.g., retention), but these are often scarce in prototypes and research projects, or too-slow to use for system optimization. We present AutoMetrics, a framework for synthesizing evaluation metrics under low-data constraints. AutoMetrics combines retrieval from MetricBank, a collection of 48 metrics we curate, with automatically generated LLM-as-a-Judge criteria informed by lightweight human feedback. These metrics are composed via regression to maximize correlation with human signal. AutoMetrics takes you from expensive measures to interpretable automatic metrics. Across 5 diverse tasks, AutoMetrics improves Kendall correlation with human ratings by up to 33.4% over LLM-as-a-Judge while requiring fewer than 100 feedback points. We show that AutoMetrics can be used as a proxy reward to equal effect as a verifiable reward. We release the full AutoMetrics toolkit and MetricBank to accelerate adaptive evaluation of LLM applications.

Kaibing Chen, Dong Shen, Hanwen Zhong et al.

In the field of multi-modal language models, the majority of methods are built on an architecture similar to LLaVA. These models use a single-layer ViT feature as a visual prompt, directly feeding it into the language models alongside textual tokens. However, when dealing with long sequences of visual signals or inputs such as videos, the self-attention mechanism of language models can lead to significant computational overhead. Additionally, using single-layer ViT features makes it challenging for large language models to perceive visual signals fully. This paper proposes an efficient multi-modal language model to minimize computational costs while enabling the model to perceive visual signals as comprehensively as possible. Our method primarily includes: (1) employing cross-attention to image-text interaction similar to Flamingo. (2) utilize hierarchical ViT features. (3) introduce the Mixture of Experts (MoE) mechanism to enhance model effectiveness. Our model achieves competitive scores on public multi-modal benchmarks and performs well in tasks such as image captioning and video captioning.

Di Xu, Xiang He, Tonghua Su et al.

Deep neural network (DNN) partition is a research problem that involves splitting a DNN into multiple parts and offloading them to specific locations. Because of the recent advancement in multi-access edge computing and edge intelligence, DNN partition has been considered as a powerful tool for improving DNN inference performance when the computing resources of edge and end devices are limited and the remote transmission of data from these devices to clouds is costly. This paper provides a comprehensive survey on the recent advances and challenges in DNN partition approaches over the cloud, edge, and end devices based on a detailed literature collection. We review how DNN partition works in various application scenarios, and provide a unified mathematical model of the DNN partition problem. We developed a five-dimensional classification framework for DNN partition approaches, consisting of deployment locations, partition granularity, partition constraints, optimization objectives, and optimization algorithms. Each existing DNN partition approache can be perfectly defined in this framework by instantiating each dimension into specific values. In addition, we suggest a set of metrics for comparing and evaluating the DNN partition approaches. Based on this, we identify and discuss research challenges that have not yet been investigated or fully addressed. We hope that this work helps DNN partition researchers by highlighting significant future research directions in this domain.

Shizhou Zhang, Xueqiang Lv, Yinghui Xing et al.

Current methods for incremental object detection (IOD) primarily rely on Faster R-CNN or DETR series detectors; however, these approaches do not accommodate the real-time YOLO detection frameworks. In this paper, we first identify three primary types of knowledge conflicts that contribute to catastrophic forgetting in YOLO-based incremental detectors: foreground-background confusion, parameter interference, and misaligned knowledge distillation. Subsequently, we introduce YOLO-IOD, a real-time Incremental Object Detection (IOD) framework that is constructed upon the pretrained YOLO-World model, facilitating incremental learning via a stage-wise parameter-efficient fine-tuning process. Specifically, YOLO-IOD encompasses three principal components: 1) Conflict-Aware Pseudo-Label Refinement (CPR), which mitigates the foreground-background confusion by leveraging the confidence levels of pseudo labels and identifying potential objects relevant to future tasks. 2) Importancebased Kernel Selection (IKS), which identifies and updates the pivotal convolution kernels pertinent to the current task during the current learning stage. 3) Cross-Stage Asymmetric Knowledge Distillation (CAKD), which addresses the misaligned knowledge distillation conflict by transmitting the features of the student target detector through the detection heads of both the previous and current teacher detectors, thereby facilitating asymmetric distillation between existing and newly introduced categories. We further introduce LoCo COCO, a more realistic benchmark that eliminates data leakage across stages. Experiments on both conventional and LoCo COCO benchmarks show that YOLO-IOD achieves superior performance with minimal forgetting.

Haiyang Ying, Baowei Jiang, Jinzhi Zhang et al.

This paper proposes a method for fast scene radiance field reconstruction with strong novel view synthesis performance and convenient scene editing functionality. The key idea is to fully utilize semantic parsing and primitive extraction for constraining and accelerating the radiance field reconstruction process. To fulfill this goal, a primitive-aware hybrid rendering strategy was proposed to enjoy the best of both volumetric and primitive rendering. We further contribute a reconstruction pipeline conducts primitive parsing and radiance field learning iteratively for each input frame which successfully fuses semantic, primitive, and radiance information into a single framework. Extensive evaluations demonstrate the fast reconstruction ability, high rendering quality, and convenient editing functionality of our method.

Di Xu, Hengjie Liu, Yang Yang et al.

Accelerated dynamic volumetric magnetic resonance imaging (4DMRI) is essential for applications relying on motion resolution. Existing 4DMRI produces acceptable artifacts of averaged breathing phases, which can blur and misrepresent instantaneous dynamic information. Recovery of such information requires a new paradigm to reconstruct extremely undersampled non-Cartesian k-space data. We propose B-FIRE, a binning-free diffusion implicit neural representation framework for hyper-accelerated MR reconstruction capable of reflecting instantaneous 3D abdominal anatomy. B-FIRE employs a CNN-INR encoder-decoder backbone optimized using diffusion with a comprehensive loss that enforces image-domain fidelity and frequency-aware constraints. Motion binned image pairs were used as training references, while inference was performed on binning-free undersampled data. Experiments were conducted on a T1-weighted StarVIBE liver MRI cohort, with accelerations ranging from 8 spokes per frame (RV8) to RV1. B-FIRE was compared against direct NuFFT, GRASP-CS, and an unrolled CNN method. Reconstruction fidelity, motion trajectory consistency, and inference latency were evaluated.

Di Xu, Hengjie Liu, Dan Ruan et al.

Dynamic magnetic resonance imaging (DMRI) is an effective imaging tool for diagnosis tasks that require motion tracking of a certain anatomy. To speed up DMRI acquisition, k-space measurements are commonly undersampled along spatial or spatial-temporal domains. The difficulty of recovering useful information increases with increasing undersampling ratios. Compress sensing was invented for this purpose and has become the most popular method until deep learning (DL) based DMRI reconstruction methods emerged in the past decade. Nevertheless, existing DL networks are still limited in long-range sequential dependency understanding and computational efficiency and are not fully automated. Considering the success of Transformers positional embedding and "swin window" self-attention mechanism in the vision community, especially natural video understanding, we hereby propose a novel architecture named Reconstruction Swin Transformer (RST) for 4D MRI. RST inherits the backbone design of the Video Swin Transformer with a novel reconstruction head introduced to restore pixel-wise intensity. A convolution network called SADXNet is used for rapid initialization of 2D MR frames before RST learning to effectively reduce the model complexity, GPU hardware demand, and training time. Experimental results in the cardiac 4D MR dataset further substantiate the superiority of RST, achieving the lowest RMSE of 0.0286 +/- 0.0199 and 1 - SSIM of 0.0872 +/- 0.0783 on 9 times accelerated validation sequences.

Xueqiang Lv, Shizhou Zhang, Yinghui Xing et al.

Open-world object detection (OWOD) requires incrementally detecting known categories while reliably identifying unknown objects. Existing methods primarily focus on improving unknown recall, yet overlook interpretability, often leading to known-unknown confusion and reduced prediction reliability. This paper aims to make the entire OWOD framework interpretable, enabling the detector to truly "knowing the unknown". To this end, we propose a concept-driven InterPretable OWOD framework(IPOW) by introducing a Concept Decomposition Model (CDM) for OWOD, which explicitly decomposes the coupled RoI features in Faster R-CNN into discriminative, shared, and background concepts. Discriminative concepts identify the most discriminative features to enlarge the distances between known categories, while shared and background concepts, due to their strong generalization ability, can be readily transferred to detect unknown categories. Leveraging the interpretable framework, we identify that known-unknown confusion arises when unknown objects fall into the discriminative space of known classes. To address this, we propose Concept-Guided Rectification (CGR) to further resolve such confusion. Extensive experiments show that IPOW significantly improves unknown recall while mitigating confusion, and provides concept-level interpretability for both known and unknown predictions.

Qirui Wu, Shizhou Zhang, De Cheng et al.

Catastrophic forgetting is a critical chanllenge for incremental object detection (IOD). Most existing methods treat the detector monolithically, relying on instance replay or knowledge distillation without analyzing component-specific forgetting. Through dissection of Faster R-CNN, we reveal a key insight: Catastrophic forgetting is predominantly localized to the RoI Head classifier, while regressors retain robustness across incremental stages. This finding challenges conventional assumptions, motivating us to develop a framework termed NSGP-RePRE. Regional Prototype Replay (RePRE) mitigates classifier forgetting via replay of two types of prototypes: coarse prototypes represent class-wise semantic centers of RoI features, while fine-grained prototypes model intra-class variations. Null Space Gradient Projection (NSGP) is further introduced to eliminate prototype-feature misalignment by updating the feature extractor in directions orthogonal to subspace of old inputs via gradient projection, aligning RePRE with incremental learning dynamics. Our simple yet effective design allows NSGP-RePRE to achieve state-of-the-art performance on the Pascal VOC and MS COCO datasets under various settings. Our work not only advances IOD methodology but also provide pivotal insights for catastrophic forgetting mitigation in IOD. Code is available at \href{https://github.com/fanrena/NSGP-RePRE}{https://github.com/fanrena/NSGP-RePRE} .

Peng Xia, Di Xu, Ming Hu et al.

Long-tailed multi-label visual recognition (LTML) task is a highly challenging task due to the label co-occurrence and imbalanced data distribution. In this work, we propose a unified framework for LTML, namely prompt tuning with class-specific embedding loss (LMPT), capturing the semantic feature interactions between categories by combining text and image modality data and improving the performance synchronously on both head and tail classes. Specifically, LMPT introduces the embedding loss function with class-aware soft margin and re-weighting to learn class-specific contexts with the benefit of textual descriptions (captions), which could help establish semantic relationships between classes, especially between the head and tail classes. Furthermore, taking into account the class imbalance, the distribution-balanced loss is adopted as the classification loss function to further improve the performance on the tail classes without compromising head classes. Extensive experiments are conducted on VOC-LT and COCO-LT datasets, which demonstrates that our method significantly surpasses the previous state-of-the-art methods and zero-shot CLIP in LTML. Our codes are fully public at https://github.com/richard-peng-xia/LMPT.

Gang Li, Di Xu, Xing Cheng et al.

Although vision Transformers have achieved excellent performance as backbone models in many vision tasks, most of them intend to capture global relations of all tokens in an image or a window, which disrupts the inherent spatial and local correlations between patches in 2D structure. In this paper, we introduce a simple vision Transformer named SimViT, to incorporate spatial structure and local information into the vision Transformers. Specifically, we introduce Multi-head Central Self-Attention(MCSA) instead of conventional Multi-head Self-Attention to capture highly local relations. The introduction of sliding windows facilitates the capture of spatial structure. Meanwhile, SimViT extracts multi-scale hierarchical features from different layers for dense prediction tasks. Extensive experiments show the SimViT is effective and efficient as a general-purpose backbone model for various image processing tasks. Especially, our SimViT-Micro only needs 3.3M parameters to achieve 71.1% top-1 accuracy on ImageNet-1k dataset, which is the smallest size vision Transformer model by now. Our code will be available in https://github.com/ucasligang/SimViT.

Hengjie Liu, Yimeng Dou, Di Xu et al.

In this paper, we present our submission to the LUMIR25 task of Learn2Reg 2025, which ranked 1st overall on the test set. Extended from LUMIR24, this year's task focuses on zero-shot registration under domain shifts (e.g., high-field MRI, pathological brains, and various MRI contrasts), while the training data comprises only in-domain T1-weighted brain MRI. We start with a meticulous analysis of LUMIR24 winners to identify the main contributors to strong monomodal registration performance. We highlight the importance of registration-specific inductive biases, including multi-resolution pyramids, inverse and group consistency, topological preservation or diffeomorphism, and correlation-based correspondence establishment. To further generalize to diverse contrasts, we employ three simple but effective strategies: (i) a multimodal loss based on the modality-independent neighborhood descriptor (MIND), (ii) intensity randomization for unseen contrast augmentation, and (iii) lightweight instance-specific optimization (ISO) on feature encoders at inference time. On the validation set, the proposed approach substantially improves T1-T2 registration accuracy, demonstrating robust cross-contrast generalization without relying on explicit image synthesis. These results suggest a practical step toward a registration foundation model that can leverage a single training domain yet remain robust across domain shifts.

Zhanfeng Liao, Yuelang Xu, Zhe Li et al.

Creating high-fidelity 3D head avatars has always been a research hotspot, but it remains a great challenge under lightweight sparse view setups. In this paper, we propose HHAvatar represented by controllable 3D Gaussians for high-fidelity head avatar with dynamic hair modeling. We first use 3D Gaussians to represent the appearance of the head, and then jointly optimize neutral 3D Gaussians and a fully learned MLP-based deformation field to capture complex expressions. The two parts benefit each other, thereby our method can model fine-grained dynamic details while ensuring expression accuracy. Furthermore, we devise a well-designed geometry-guided initialization strategy based on implicit SDF and Deep Marching Tetrahedra for the stability and convergence of the training procedure. To address the problem of dynamic hair modeling, we introduce a hybrid head model into our avatar representation based Gaussian Head Avatar and a training method that considers timing information and an occlusion perception module to model the non-rigid motion of hair. Experiments show that our approach outperforms other state-of-the-art sparse-view methods, achieving ultra high-fidelity rendering quality at 2K resolution even under exaggerated expressions and driving hairs reasonably with the motion of the head

Jakob Schwerter, Loreen Sabel, Judith Bose et al.

STEM dropout rates remain high at universities, particularly in computer science programs with theory-intensive courses. Digital learning environments now capture rich behavioral data that could help identify struggling students early, yet the generalizability of data-driven prediction models across courses and institutions remains uncertain. Guided by self-regulated learning (SRL) theory, this study analyzed multimodal digital-trace data from three undergraduate theoretical computer science courses (N1 = 137, N2 = 104, N3 = 148) at two universities. Weekly SRL-aligned digital-trace indicators were modeled using Elastic Net, Random Forest, and XGBoost to evaluate predictive performance over time and across settings, and model calibration both within and across courses. Early prediction of at-risk students was feasible, with SRL-related behaviors such as time management, effort regulation, and sustained engagement emerging as key predictors. While Random Forest achieved the highest in-sample accuracy, Elastic Net generalized more robustly across contexts. Out-of-sample accuracy and calibration declined between institutions with different base rates, underscoring the contextual nature of predictive analytics in higher education. These findings suggest that digital learning traces enable early identification of at-risk students within courses, but generalizing predictive models beyond their original context requires caution, particularly if the at-risk rates differ between contexts.

Kwai Keye Team, Biao Yang, Bin Wen et al.

While Multimodal Large Language Models (MLLMs) demonstrate remarkable capabilities on static images, they often fall short in comprehending dynamic, information-dense short-form videos, a dominant medium in today's digital landscape. To bridge this gap, we introduce \textbf{Kwai Keye-VL}, an 8-billion-parameter multimodal foundation model engineered for leading-edge performance in short-video understanding while maintaining robust general-purpose vision-language abilities. The development of Keye-VL rests on two core pillars: a massive, high-quality dataset exceeding 600 billion tokens with a strong emphasis on video, and an innovative training recipe. This recipe features a four-stage pre-training process for solid vision-language alignment, followed by a meticulous two-phase post-training process. The first post-training stage enhances foundational capabilities like instruction following, while the second phase focuses on stimulating advanced reasoning. In this second phase, a key innovation is our five-mode ``cold-start'' data mixture, which includes ``thinking'', ``non-thinking'', ``auto-think'', ``think with image'', and high-quality video data. This mixture teaches the model to decide when and how to reason. Subsequent reinforcement learning (RL) and alignment steps further enhance these reasoning capabilities and correct abnormal model behaviors, such as repetitive outputs. To validate our approach, we conduct extensive evaluations, showing that Keye-VL achieves state-of-the-art results on public video benchmarks and remains highly competitive on general image-based tasks (Figure 1). Furthermore, we develop and release the \textbf{KC-MMBench}, a new benchmark tailored for real-world short-video scenarios, where Keye-VL shows a significant advantage.

Ecem Sogancioglu, Bram van Ginneken, Finn Behrendt et al.

Pulmonary nodules may be an early manifestation of lung cancer, the leading cause of cancer-related deaths among both men and women. Numerous studies have established that deep learning methods can yield high-performance levels in the detection of lung nodules in chest X-rays. However, the lack of gold-standard public datasets slows down the progression of the research and prevents benchmarking of methods for this task. To address this, we organized a public research challenge, NODE21, aimed at the detection and generation of lung nodules in chest X-rays. While the detection track assesses state-of-the-art nodule detection systems, the generation track determines the utility of nodule generation algorithms to augment training data and hence improve the performance of the detection systems. This paper summarizes the results of the NODE21 challenge and performs extensive additional experiments to examine the impact of the synthetically generated nodule training images on the detection algorithm performance.

Likun Li, Haoqi Zeng, Changpeng Yang et al.

The objective of personalization and stylization in text-to-image is to instruct a pre-trained diffusion model to analyze new concepts introduced by users and incorporate them into expected styles. Recently, parameter-efficient fine-tuning (PEFT) approaches have been widely adopted to address this task and have greatly propelled the development of this field. Despite their popularity, existing efficient fine-tuning methods still struggle to achieve effective personalization and stylization in T2I generation. To address this issue, we propose block-wise Low-Rank Adaptation (LoRA) to perform fine-grained fine-tuning for different blocks of SD, which can generate images faithful to input prompts and target identity and also with desired style. Extensive experiments demonstrate the effectiveness of the proposed method.

Biao Yang, Bin Wen, Boyang Ding et al.

In recent years, the development of Large Language Models (LLMs) has significantly advanced, extending their capabilities to multimodal tasks through Multimodal Large Language Models (MLLMs). However, video understanding remains a challenging area due to the dynamic and information-dense nature of videos. Existing models struggle with the trade-off between spatial resolution and temporal coverage when processing video content. We present Keye-VL-1.5, which addresses fundamental challenges in video comprehension through three key innovations. First, we introduce a novel Slow-Fast video encoding strategy that dynamically allocates computational resources based on inter-frame similarity, processing key frames with significant visual changes at higher resolution (Slow pathway) while handling relatively static frames with increased temporal coverage at lower resolution (Fast pathway). Second, we implement a progressive four-stage pre-training methodology that systematically extends the model's context length from 8K to 128K tokens, enabling processing of longer videos and more complex visual content. Third, we develop a comprehensive post-training pipeline focusing on reasoning enhancement and human preference alignment, incorporating a 5-step chain-of-thought data construction process, iterative GSPO-based reinforcement learning with progressive prompt hinting for difficult cases, and alignment training. Through extensive evaluation on public benchmarks and rigorous internal human assessment, Keye-VL-1.5 demonstrates significant improvements over existing models, particularly excelling in video understanding tasks while maintaining competitive performance on general multimodal benchmarks.

Ashish Shenoy, Yichao Lu, Srihari Jayakumar et al.

We introduce Lumos, the first end-to-end multimodal question-answering system with text understanding capabilities. At the core of Lumos is a Scene Text Recognition (STR) component that extracts text from first person point-of-view images, the output of which is used to augment input to a Multimodal Large Language Model (MM-LLM). While building Lumos, we encountered numerous challenges related to STR quality, overall latency, and model inference. In this paper, we delve into those challenges, and discuss the system architecture, design choices, and modeling techniques employed to overcome these obstacles. We also provide a comprehensive evaluation for each component, showcasing high quality and efficiency.

Di Xu, Xin Miao, Hengjie Liu et al.

Purpose: 4D MRI with high spatiotemporal resolution is desired for image-guided liver radiotherapy. Acquiring densely sampling k-space data is time-consuming. Accelerated acquisition with sparse samples is desirable but often causes degraded image quality or long reconstruction time. We propose the Reconstruct Paired Conditional Generative Adversarial Network (Re-Con-GAN) to shorten the 4D MRI reconstruction time while maintaining the reconstruction quality. Methods: Patients who underwent free-breathing liver 4D MRI were included in the study. Fully- and retrospectively under-sampled data at 3, 6 and 10 times (3x, 6x and 10x) were first reconstructed using the nuFFT algorithm. Re-Con-GAN then trained input and output in pairs. Three types of networks, ResNet9, UNet and reconstruction swin transformer, were explored as generators. PatchGAN was selected as the discriminator. Re-Con-GAN processed the data (3D+t) as temporal slices (2D+t). A total of 48 patients with 12332 temporal slices were split into training (37 patients with 10721 slices) and test (11 patients with 1611 slices). Results: Re-Con-GAN consistently achieved comparable/better PSNR, SSIM, and RMSE scores compared to CS/UNet models. The inference time of Re-Con-GAN, UNet and CS are 0.15s, 0.16s, and 120s. The GTV detection task showed that Re-Con-GAN and CS, compared to UNet, better improved the dice score (3x Re-Con-GAN 80.98%; 3x CS 80.74%; 3x UNet 79.88%) of unprocessed under-sampled images (3x 69.61%). Conclusion: A generative network with adversarial training is proposed with promising and efficient reconstruction results demonstrated on an in-house dataset. The rapid and qualitative reconstruction of 4D liver MR has the potential to facilitate online adaptive MR-guided radiotherapy for liver cancer.

Haozhe Jia, Yan Li, Hengfei Cui et al.

In this work, we focus on exploring explicit fine-grained control of generative facial image editing, all while generating faithful facial appearances and consistent semantic details, which however, is quite challenging and has not been extensively explored, especially under an one-shot scenario. We identify the key challenge as the exploration of disentangled conditional control between high-level semantics and explicit parameters (e.g., 3DMM) in the generation process, and accordingly propose a novel diffusion-based editing framework, named DisControlFace. Specifically, we leverage a Diffusion Autoencoder (Diff-AE) as the semantic reconstruction backbone. To enable explicit face editing, we construct an Exp-FaceNet that is compatible with Diff-AE to generate spatial-wise explicit control conditions based on estimated 3DMM parameters. Different from current diffusion-based editing methods that train the whole conditional generative model from scratch, we freeze the pre-trained weights of the Diff-AE to maintain its semantically deterministic conditioning capability and accordingly propose a random semantic masking (RSM) strategy to effectively achieve an independent training of Exp-FaceNet. This setting endows the model with disentangled face control meanwhile reducing semantic information shift in editing. Our model can be trained using 2D in-the-wild portrait images without requiring 3D or video data and perform robust editing on any new facial image through a simple one-shot fine-tuning. Comprehensive experiments demonstrate that DisControlFace can generate realistic facial images with better editing accuracy and identity preservation over state-of-the-art methods. Project page: https://discontrolface.github.io/

Di Xu, Xin Miao, Hengjie Liu et al.

Abstract Purpose: High-quality 4D MRI requires an impractically long scanning time for dense k-space signal acquisition covering all respiratory phases. Accelerated sparse sampling followed by reconstruction enhancement is desired but often results in degraded image quality and long reconstruction time. We hereby propose the chained iterative reconstruction network (CIRNet) for efficient sparse-sampling reconstruction while maintaining clinically deployable quality. Methods: CIRNet adopts the denoising diffusion probabilistic framework to condition the image reconstruction through a stochastic iterative denoising process. During training, a forward Markovian diffusion process is designed to gradually add Gaussian noise to the densely sampled ground truth (GT), while CIRNet is optimized to iteratively reverse the Markovian process from the forward outputs. At the inference stage, CIRNet performs the reverse process solely to recover signals from noise, conditioned upon the undersampled input. CIRNet processed the 4D data (3D+t) as temporal slices (2D+t). The proposed framework is evaluated on a data cohort consisting of 48 patients (12332 temporal slices) who underwent free-breathing liver 4D MRI. 3-, 6-, 10-, 20- and 30-times acceleration were examined with a retrospective random undersampling scheme. Compressed sensing (CS) reconstruction with a spatiotemporal constraint and a recently proposed deep network, Re-Con-GAN, are selected as baselines. Results: CIRNet consistently achieved superior performance compared to CS and Re-Con-GAN. The inference time of CIRNet, CS, and Re-Con-GAN are 11s, 120s, and 0.15s. Conclusion: A novel framework, CIRNet, is presented. CIRNet maintains useable image quality for acceleration up to 30 times, significantly reducing the burden of 4DMRI.

Di Xu, Yang Yang, Hengjie Liu et al.

Computed tomography (CT) provides high spatial resolution visualization of 3D structures for scientific and clinical applications. Traditional analytical/iterative CT reconstruction algorithms require hundreds of angular data samplings, a condition that may not be met in practice due to physical and mechanical limitations. Sparse view CT reconstruction has been proposed using constrained optimization and machine learning methods with varying success, less so for ultra-sparse view CT reconstruction with one to two views. Neural radiance field (NeRF) is a powerful tool for reconstructing and rendering 3D natural scenes from sparse views, but its direct application to 3D medical image reconstruction has been minimally successful due to the differences between optical and X-ray photon transportation. Here, we develop a novel TomoGRAF framework incorporating the unique X-ray transportation physics to reconstruct high-quality 3D volumes using ultra-sparse projections without prior. TomoGRAF captures the CT imaging geometry, simulates the X-ray casting and tracing process, and penalizes the difference between simulated and ground truth CT sub-volume during training. We evaluated the performance of TomoGRAF on an unseen dataset of distinct imaging characteristics from the training data and demonstrated a vast leap in performance compared with state-of-the-art deep learning and NeRF methods. TomoGRAF provides the first generalizable solution for image-guided radiotherapy and interventional radiology applications, where only one or a few X-ray views are available, but 3D volumetric information is desired.

Lingfeng He, De Cheng, Di Xu et al.

Continual learning (CL) aims to equip models with the ability to learn from a stream of tasks without forgetting previous knowledge. With the progress of vision-language models like Contrastive Language-Image Pre-training (CLIP), their promise for CL has attracted increasing attention due to their strong generalizability. However, the potential of rich textual semantic priors in CLIP in addressing the stability-plasticity dilemma remains underexplored. During backbone training, most approaches transfer past knowledge without considering semantic relevance, leading to interference from unrelated tasks that disrupt the balance between stability and plasticity. Besides, while text-based classifiers provide strong generalization, they suffer from limited plasticity due to the inherent modality gap in CLIP. Visual classifiers help bridge this gap, but their prototypes lack rich and precise semantics. To address these challenges, we propose Semantic-Enriched Continual Adaptation (SECA), a unified framework that harnesses the anti-forgetting and structured nature of textual priors to guide semantic-aware knowledge transfer in the backbone and reinforce the semantic structure of the visual classifier. Specifically, a Semantic-Guided Adaptive Knowledge Transfer (SG-AKT) module is proposed to assess new images' relevance to diverse historical visual knowledge via textual cues, and aggregate relevant knowledge in an instance-adaptive manner as distillation signals. Moreover, a Semantic-Enhanced Visual Prototype Refinement (SE-VPR) module is introduced to refine visual prototypes using inter-class semantic relations captured in class-wise textual embeddings. Extensive experiments on multiple benchmarks validate the effectiveness of our approach.

Di Xu

Efficiently acquired and precisely reconstructed imaging are crucial to the success of modern radiation therapy (RT). Computed tomography (CT) and magnetic resonance imaging (MRI) are two common modalities for providing RT treatment planning and delivery guidance/monitoring. In recent decades, artificial intelligence (AI) has emerged as a powerful and widely adopted technique across various fields, valued for its efficiency and convenience enabled by implicit function definition and data-driven feature representation learning. Here, we present a series of AI-driven medical imaging reconstruction frameworks for enhanced radiotherapy, designed to improve CT image reconstruction quality and speed, refine dual-energy CT (DECT) multi-material decomposition (MMD), and significantly accelerate 4D MRI acquisition.

Di Xu, Hengjie Liu, Xin Miao et al.

The scanning time for a fully sampled MRI can be undesirably lengthy. Compressed sensing has been developed to minimize image artifacts in accelerated scans, but the required iterative reconstruction is computationally complex and difficult to generalize on new cases. Image-domain-based deep learning methods (e.g., convolutional neural networks) emerged as a faster alternative but face challenges in modeling continuous k-space, a problem amplified with non-Cartesian sampling commonly used in accelerated acquisition. In comparison, implicit neural representations can model continuous signals in the frequency domain and thus are compatible with arbitrary k-space sampling patterns. The current study develops a novel generative-adversarially trained implicit neural representations (k-GINR) for de novo undersampled non-Cartesian k-space reconstruction. k-GINR consists of two stages: 1) supervised training on an existing patient cohort; 2) self-supervised patient-specific optimization. In stage 1, the network is trained with the generative-adversarial network on diverse patients of the same anatomical region supervised by fully sampled acquisition. In stage 2, undersampled k-space data of individual patients is used to tailor the prior-embedded network for patient-specific optimization. The UCSF StarVIBE T1-weighted liver dataset was evaluated on the proposed framework. k-GINR is compared with an image-domain deep learning method, Deep Cascade CNN, and a compressed sensing method. k-GINR consistently outperformed the baselines with a larger performance advantage observed at very high accelerations (e.g., 20 times). k-GINR offers great value for direct non-Cartesian k-space reconstruction for new incoming patients across a wide range of accelerations liver anatomy.

Xuanchen Li, Yuhao Cheng, Xingyu Ren et al.

4D head capture aims to generate dynamic topological meshes and corresponding texture maps from videos, which is widely utilized in movies and games for its ability to simulate facial muscle movements and recover dynamic textures in pore-squeezing. The industry often adopts the method involving multi-view stereo and non-rigid alignment. However, this approach is prone to errors and heavily reliant on time-consuming manual processing by artists. To simplify this process, we propose Topo4D, a novel framework for automatic geometry and texture generation, which optimizes densely aligned 4D heads and 8K texture maps directly from calibrated multi-view time-series images. Specifically, we first represent the time-series faces as a set of dynamic 3D Gaussians with fixed topology in which the Gaussian centers are bound to the mesh vertices. Afterward, we perform alternative geometry and texture optimization frame-by-frame for high-quality geometry and texture learning while maintaining temporal topology stability. Finally, we can extract dynamic facial meshes in regular wiring arrangement and high-fidelity textures with pore-level details from the learned Gaussians. Extensive experiments show that our method achieves superior results than the current SOTA face reconstruction methods both in the quality of meshes and textures. Project page: https://xuanchenli.github.io/Topo4D/.

Di Xu, Yang Zhao, Xiang Hao et al.

We introduce a novel dataset consisting of images depicting pink eggs that have been identified as Pomacea canaliculata eggs, accompanied by corresponding bounding box annotations. The purpose of this dataset is to aid researchers in the analysis of the spread of Pomacea canaliculata species by utilizing deep learning techniques, as well as supporting other investigative pursuits that require visual data pertaining to the eggs of Pomacea canaliculata. It is worth noting, however, that the identity of the eggs in question is not definitively established, as other species within the same taxonomic family have been observed to lay similar-looking eggs in regions of the Americas. Therefore, a crucial prerequisite to any decision regarding the elimination of these eggs would be to establish with certainty whether they are exclusively attributable to invasive Pomacea canaliculata or if other species are also involved. The dataset is available at https://www.kaggle.com/datasets/deeshenzhen/pinkeggs

Jillian M. Clements, Di Xu, Nooshin Yousefi et al.

Machine learning plays an essential role in preventing financial losses in the banking industry. Perhaps the most pertinent prediction task that can result in billions of dollars in losses each year is the assessment of credit risk (i.e., the risk of default on debt). Today, much of the gains from machine learning to predict credit risk are driven by gradient boosted decision tree models. However, these gains begin to plateau without the addition of expensive new data sources or highly engineered features. In this paper, we present our attempts to create a novel approach to assessing credit risk using deep learning that does not rely on new model inputs. We propose a new credit card transaction sampling technique to use with deep recurrent and causal convolution-based neural networks that exploits long historical sequences of financial data without costly resource requirements. We show that our sequential deep learning approach using a temporal convolutional network outperformed the benchmark non-sequential tree-based model, achieving significant financial savings and earlier detection of credit risk. We also demonstrate the potential for our approach to be used in a production environment, where our sampling technique allows for sequences to be stored efficiently in memory and used for fast online learning and inference.

Di Xu, Zhen Li, Yanning Zhang et al.

This paper presents an illumination estimation method for virtual objects in real environment by learning. While previous works tackled this problem by reconstructing high dynamic range (HDR) environment maps or the corresponding spherical harmonics, we do not seek to recover the lighting environment of the entire scene. Given a single RGB image, our method directly infers the relit virtual object by transferring the illumination features extracted from planar surfaces in the scene to the desired geometries. Compared to previous works, our approach is more robust as it works in both indoor and outdoor environments with spatially-varying illumination. Experiments and evaluation results show that our approach outperforms the state-of-the-art quantitatively and qualitatively, achieving realistic augmented experience.

Dmitry Efimov, Di Xu, Luyang Kong et al.

Generative Adversarial Networks (GANs) became very popular for generation of realistically looking images. In this paper, we propose to use GANs to synthesize artificial financial data for research and benchmarking purposes. We test this approach on three American Express datasets, and show that properly trained GANs can replicate these datasets with high fidelity. For our experiments, we define a novel type of GAN, and suggest methods for data preprocessing that allow good training and testing performance of GANs. We also discuss methods for evaluating the quality of generated data, and their comparison with the original real data.

Di Xu, Tianhang Long, Junbin Gao

Massive volumes of high-dimensional data that evolves over time is continuously collected by contemporary information processing systems, which brings up the problem of organizing this data into clusters, i.e. achieve the purpose of dimensional deduction, and meanwhile learning its temporal evolution patterns. In this paper, a framework for evolutionary subspace clustering, referred to as LSTM-ESCM, is introduced, which aims at clustering a set of evolving high-dimensional data points that lie in a union of low-dimensional evolving subspaces. In order to obtain the parsimonious data representation at each time step, we propose to exploit the so-called self-expressive trait of the data at each time point. At the same time, LSTM networks are implemented to extract the inherited temporal patterns behind data in an overall time frame. An efficient algorithm has been proposed based on MATLAB. Next, experiments are carried out on real-world datasets to demonstrate the effectiveness of our proposed approach. And the results show that the suggested algorithm dramatically outperforms other known similar approaches in terms of both run time and accuracy.

Di Xu, Manjing Fang, Xia Hong et al.

A general framework of least squares support vector machine with low rank kernels, referred to as LR-LSSVM, is introduced in this paper. The special structure of low rank kernels with a controlled model size brings sparsity as well as computational efficiency to the proposed model. Meanwhile, a two-step optimization algorithm with three different criteria is proposed and various experiments are carried out using the example of the so-call robust RBF kernel to validate the model. The experiment results show that the performance of the proposed algorithm is comparable or superior to several existing kernel machines.