AAAI 2025 Papers

Monocular camera calibration is a key precondition for numerous 3D vision applications. Despite considerable advancements, existing methods often hinge on specific assumptions and struggle to generalize across varied real-world scenarios, and the performance is limited by insufficient training data. Recently, diffusion models trained on expansive datasets have been confirmed to maintain the capability to generate diverse, high-quality images. This success suggests a strong potential of the models to effectively understand varied visual information. In this work, we leverage the comprehensive visual knowledge embedded in pre-trained diffusion models to enable more robust and accurate monocular camera intrinsic estimation. Specifically, we reformulate the problem of estimating the four degrees of freedom (4-DoF) of camera intrinsic parameters as a dense incident map generation task. The map details the angle of incidence for each pixel in the RGB image, and its format aligns well with the paradigm of diffusion models. The camera intrinsic then can be derived from the incident map with a simple non-learning RANSAC algorithm during inference. Moreover, to further enhance the performance, we jointly estimate a depth map to provide extra geometric information for the incident map estimation. Extensive experiments on multiple testing datasets demonstrates that our model achieves state-of-the-art performance, gaining up to a 40% reduction in prediction errors. Besides, the experiments also show that the precise camera intrinsic and depth maps estimated by our pipeline can greatly benefit practical applications such as 3D reconstruction from a single in-the-wild image.

Visual language models like Contrastive Language-Image Pretraining (CLIP) have shown impressive performance in analyzing natural images with language information. However, these models often encounter challenges when applied to specialized domains such as remote sensing due to the limited availability of image-text pairs for training. To tackle this issue, we introduce DiffCLIP, a novel framework that extends CLIP to effectively convey comprehensive language-driven semantic information for accurate classification of high-dimensional multimodal remote sensing images. DiffCLIP is a few-shot learning method that leverages unlabeled images for pretraining. It employs unsupervised mask diffusion learning to capture the distribution of diverse modalities without requiring labels. The modality-shared image encoder maps multimodal data into a unified subspace, extracting shared features with consistent parameters across modalities. A well-trained image encoder further enhances learning by aligning visual representations with class-label text information from CLIP. By integrating these approaches, DiffCLIP significantly boosts CLIP performance using a minimal number of image-text pairs. We evaluate DiffCLIP on widely used high-dimensional multimodal datasets, demonstrating its effectiveness in addressing few-shot annotated classification tasks. DiffCLIP achieves an overall accuracy improvement of 10.65% across three remote sensing datasets compared with CLIP, while utilizing only 2-shot image-text pairs.

We propose Diff-Shadow, a global-guided diffusion model for high-quality shadow removal. Previous transformer-based approaches can utilize global information to relate shadow and non-shadow regions but are limited in their synthesis ability and recover images with obvious boundaries. In contrast, diffusion-based methods can generate better content but they are not exempt from issues related to inconsistent illumination. In this work, we combine the advantages of diffusion models and global guidance to realize shadow-free restoration. Specifically, we propose a parallel UNets architecture: 1) the local branch performs the patch-based noise estimation in the diffusion process, and 2) the global branch recovers the low-resolution shadow-free images. A Reweight Cross Attention (RCA) module is designed to integrate global contextual information of non-shadow regions into the local branch. We further design a Global-guided Sampling Strategy (GSS) that mitigates patch boundary issues and ensures consistent illumination across shaded and unshaded regions in the recovered image. Comprehensive experiments on three publicly standard datasets ISTD, ISTD+, and SRD have demonstrated the effectiveness of Diff-Shadow. Compared to state-of-the-art methods, our method achieves a significant improvement in terms of PSNR, increasing from 32.33dB to 33.69dB on the ISTD dataset.

In recent years, diffusion modeling has shown great potential for image generation and editing. Beyond single-model approaches, various drawing workflows now exist to handle diverse drawing tasks. However, few solutions effectively identify user intentions through dialogue and progressively complete drawings. We introduce DialogDraw, which facilitates image generation and editing through continuous dialogue interaction. DialogDraw enables users to create and refine drawings using natural language and integrates with numerous open-source drawing workflows and models. The system accurately recognizes intentions and extracts user inputs via parameterization, adapts to various drawing function parameters, and provides an intuitive interaction mode. It effectively executes user instructions, supports dozens of image generation and editing methods, and offers robust scalability. Moreover, we employ SFT and RLHF to iterate the Intention Recognition and Parameter Extraction Model (IRPEM). To evaluate DialogDraw's functionality, we propose DrawnConvos, a dataset rich in drawing functions and command dialogue data collected from the open-source community. Our evaluation demonstrates that DialogDraw excels in command compliance, identifying and adapting to user drawing intentions, thereby proving the effectiveness of our method.

Despite recent advances in UNet-based image editing, methods for shape-aware object editing in high-resolution images are still lacking. Compared to UNet, Diffusion Transformers (DiT) demonstrate superior capabilities to effectively capture the long-range dependencies among patches, leading to higher-quality image generation. In this paper, we propose DiT4Edit, the first Diffusion Transformer-based image editing framework. Specifically, DiT4Edit uses the DPM-Solver inversion algorithm to obtain the inverted latents, reducing the number of steps compared to the DDIM inversion algorithm commonly used in UNet-based frameworks. Additionally, we design unified attention control and patch merging, tailored for transformer computation streams. This integration allows our framework to generate higher-quality edited images faster. Our design leverages the advantages of DiT, enabling it to surpass UNet structures in image editing, especially in high-resolution and arbitrary-size images. Extensive experiments demonstrate the strong performance of DiT4Edit in various editing scenarios, highlighting the potential of diffusion transformers for image editing.

Multi-modal salient object detection (SOD) through the integration of additional data such as depth or thermal information has become a significant task in computer vision during recent years. Traditionally, the challenges of identifying salient objects in RGB, RGB-D (Depth), and RGB-T (Thermal) images are tackled separately. However, without intricate cross-modal fusion strategies, such approaches struggle to effectively integrate multi-modal information, often resulting in poorly defined object edges or overconfident inaccurate predictions. Recent studies have shown that designing a unified end-to-end framework to handle all three types of SOD tasks simultaneously is both necessary and difficult. To address this need, we propose a novel approach that treats multi-modal SOD as a conditional mask generation task utilizing diffusion models. We introduce DiMSOD, which enables the concurrent use of local (depth maps, thermal maps) and global controls (original images) within a unified model for progressive denoising and refined prediction. DiMSOD is efficient, only requiring fine-tuning of our newly introduced modules on the existing stable diffusion, which not only reduces the fine-tuning cost, making it more viable for practical use, but also enhances the integration of multi-modal conditional controls. Specifically, we have developed modules including SOD-ControlNet, Feature Adaptive Network (FAN), and Feature Injection Attention Network (FIAN) to enhance the model's performance. Extensive experiments demonstrate that DiMSOD efficiently detects salient objects across RGB, RGB-D, and RGB-T datasets, achieving superior performance compared to previous well-established methods.

Cross-modal retrieval aims to retrieve relevant data across different modalities. Driven by costly massive labeled data, existing cross-modal retrieval methods achieve encouraging results. To reduce annotation costs while maintaining performance, this paper focuses on an untouched but challenging problem, i.e., cross-modal retrieval with partial labels (PLCMR). PLCMR faces the dual challenges of annotation ambiguity and modality gap. To address these challenges, we propose a novel method termed disambiguated contrastive alignment (DiCA) for cross-modal retrieval with partial labels. Specifically, DiCA proposes a novel non-candidate boosted disambiguation learning mechanism (NBDL), which elaborately balances the trade-off between the losses on candidate and non-candidate labels that eliminate label ambiguity and narrow the modality gap. Moreover, DiCA presents an instance-prototype representation learning mechanism (IPRL) to enhance the model by further eliminating the modality gap at both the instance and prototype levels. Thanks to NBDL and IPRL, our DiCA effectively addresses the issues of annotation ambiguity and modality gap for cross-modal retrieval with partial labels. Experiments on four benchmarks validate the effectiveness of our proposed method, which demonstrates enhanced performance over existing state-of-the-art methods.

We consider the Coalition Structure Learning (CSL) problem in multi-agent systems, motivated by the existence of coalitions in many real-world systems, e.g., trading platforms and auction systems. In this problem, there is a hidden coalition structure within a set of n agents, which affects the behavior of the agents in games. Our goal is to actively design a sequence of games for the agents to play, such that observations in these games can be used to learn the hidden coalition structure. In particular, we consider the setting where in each round, we design and present a game together with a strategy profile to the agents, and receive a multiple-bit observation -- for each agent, we observe whether or not they would like to deviate from the specified strategy. We show that we can learn the coalition structure in O(log n) rounds if we are allowed to design any normal-form game, matching the information-theoretical lower bound. For practicality, we extend the result to settings where we can only choose games of a specific format, and design algorithms to learn the coalition structure in these settings. For most settings, our complexity matches the theoretical lower bound up to a constant factor.

Artificial Intelligence (AI) has rapidly transformed the medical field, necessitating significant changes in medical education to prepare healthcare professionals for future work requirements. However, the integration of AI into medical curricula has been slow and lacks standardization. In this paper, we present our work in developing a year-long postgraduate-level AI in Medicine program offered by a medical school at a public university in Singapore. Our curriculum design follows Kern's six-step approach to medical curriculum development, organized into a four-session framework. These sessions involved collaboration with hospital and university administrators, educators, industry experts, and healthcare professionals. The program is structured around three core courses: Foundational Healthcare AI, Clinical Applications of Healthcare AI, and Governance and Ethics for Healthcare AI. Each course comprises multiple modules with associated projects, emphasizing hands-on learning. The program adopts a problem-based learning approach, supported by a blended learning environment to accommodate the schedules of working healthcare professionals. Evaluations by industry experts highlight the program's potential to address critical gaps in the healthcare sector. This study contributes to the integration of AI into medical training by providing a standardized approach that can be adapted globally.

Foundation models in general domains have leveraged multimodal knowledge graphs to great effect, yet the healthcare sector lacks such comprehensive structures, presenting a significant gap in current research. Based on previous exploration with pure data-driven approaches, this proposal describes a two-stage project aiming to enhance multimodal healthcare foundation model with domain knowledge. The first stage is to construct a robust multimodal healthcare knowledge graph based on established healthcare taxonomies, such as UMLS, and enriched with data from multimodal clinical databases like MIMIC-CXR. This knowledge graph will incorporate medical images as cross-modal instances linked to healthcare terminologies, enhancing the depth and applicability of the graph. In the second stage, the knowledge graph will serve as a foundational tool in training healthcare foundation models with enhanced capabilities, particularly in reducing hallucination and managing concept ambiguity through the novel use of reinforcement learning techniques like Direct Preference Optimization (DPO). This research is expected to make significant contributions to the domain of healthcare AI by enabling more accurate, reliable, and explainable AI-driven diagnostics and interventions.

Large Language Models (LLMs) have shown promise in educational applications, but challenges such as hallucinations, lack of contextual relevance, and limited personalization impede their practical adoption. To address these issues, my research introduces MerryQuery, an LLM-powered educational agent that integrates Retrieval-Augmented Generation (RAG), rule-based content control, and Reinforcement Learning from Human Feedback (RLHF). The system features a dynamic learning profile module for adaptive personalization and a multi-step verification framework that cross-checks responses against external sources to enhance trustworthiness. A functional prototype of MerryQuery is being piloted in a real-world classroom. Preliminary results demonstrate improved response reliability and student understanding.

According to an industry survey, many people miss opportunities to apply for government subsidy programs because they do not know how to apply. People also need to search manually and check whether these programs are suitable for them. To address this issue, our study develops a new generative recommender system with both users’ information and government subsidy documents. Within our recommender system framework, we modify the existing Residual Quantization Variational Auto-Encoder (RQ-VAE) model to capture deep and abstract information from subsidy documents. Using semantic IDs generated for approximately 185,610 user click-stream histories and 240,000 documents, we train our recommender system to predict the semantic IDs of the next subsidy policy documents in which a user might be interested. In 2024, we successfully deploy our generative recommender system in Wello, a Korean Gov-Tech startup. In collaboration with the Korean government, our generative recommender system could save 7.8 million dollar, that might otherwise have gone unused due to a lack of applications. Also, Wello observed a 68% improvement in Click-Through Ratio (CTR), increasing from 41.4% in the third quarter of 2024 to 69.6% in the fourth quarter of 2024. We thus anticipate that our generative recommender system will have a significant impact on both individuals and the government.

This paper presents an experiential learning pedagogy that teaches undergraduate business management information systems students hands-on AI skills through the lens of sustainability. The learning modules aim to empower undergraduate business students to gain interest and confidence in AI knowledge, skills, and careers, to sharpen their higher order thinking abilities, and to help them gain a deeper understanding of sustainability issues. Students learn AI through developing chatbots that address pressing sustainability issues within their own communities. Results of the pilot study indicate that students have increased self-efficacy in AI, more positive attitudes towards AI learning and AI-related careers, enhanced sustainability awareness, and more confidence in their ability to innovate.

We study the problem of computing deterministic optimal policies for constrained Markov decision processes (MDPs) with continuous state and action spaces, which are widely encountered in constrained dynamical systems. Designing deterministic policy gradient methods in continuous state and action spaces is particularly challenging due to the lack of enumerable state-action pairs and the adoption of deterministic policies, hindering the application of existing policy gradient methods for constrained MDPs. To this end, we develop a deterministic policy gradient primal-dual method to find an optimal deterministic policy with non-asymptotic convergence. Specifically, we leverage regularization of the Lagrangian of the constrained MDP to propose a deterministic policy gradient primal-dual (D-PGPD) algorithm that updates the deterministic policy via a quadratic-regularized gradient ascent step and the dual variable via a quadratic-regularized gradient descent step. We prove that the primal-dual iterates of D-PGPD converge at a sub-linear rate to an optimal regularized primal-dual pair. We instantiate D-PGPD with function approximation and prove that the primal-dual iterates of D-PGPD converge at a sub-linear rate to an optimal regularized primal-dual pair, up to a function approximation error. Furthermore, we demonstrate the effectiveness of our method in two continuous control problems: robot navigation and fluid control. To the best of our knowledge, this appears to be the first work that proposes a deterministic policy search method for continuous-space constrained MDPs.

The rapidly developing Large Vision Language Models (LVLMs) still face the hallucination phenomena where the generated responses do not align with the given contexts, significantly restricting the usages of LVLMs. Most previous work detects and mitigates hallucination at the coarse-grained level or requires expensive annotation (e.g., labeling by human experts or proprietary models). To address these issues, we propose detecting and mitigating hallucinations in LVLMs via fine-grained AI feedback. The basic idea is that we generate a small-size sentence-level hallucination annotation dataset by proprietary models, whereby we train a detection model which can perform sentence-level hallucination detection. Then, we propose a detect-then-rewrite pipeline to automatically construct preference dataset for hallucination mitigation training. Furthermore, we propose differentiating the severity of hallucinations, and introducing a Hallucination Severity-Aware Direct Preference Optimization (HSA-DPO) which prioritizes the mitigation of critical hallucination in LVLMs by incorporating the severity of hallucinations into preference learning. Extensive experiments on hallucination detection and mitigation benchmarks demonstrate that our method sets a new state-of-the-art in hallucination detection on MHaluBench, surpassing GPT-4V and Gemini, and reduces the hallucination rate by 36.1% on AMBER and 76.3% on Object HalBench compared to the base model.

Convolution-based unlearnable examples (UEs) employ class-wise multiplicative convolutional noise to training samples, severely compromising model performance. This fire-new type of UEs have successfully countered all defense mechanisms against UEs. The failure of such defenses can be attributed to the absence of norm constraints on convolutional noise, leading to severe blurring of image features. To address this, we first design an Edge Pixel-based Detector (EPD) to identify convolution-based UEs. Upon detection of them, we propose the first defense scheme against convolution-based UEs, COrrupting these samples via random matrix multiplication by employing bilinear INterpolation (COIN) such that disrupting the distribution of class-wise multiplicative noise. To evaluate the generalization of our proposed COIN, we newly design two convolution-based UEs called VUDA and HUDA to expand the scope of convolution-based UEs. Extensive experiments demonstrate the effectiveness of detection scheme EPD and that our defense COIN outperforms 11 state-of-the-art (SOTA) defenses, achieving a significant improvement on the CIFAR and ImageNet datasets.

Beginner musicians often struggle to identify specific errors in their performances, such as playing incorrect notes or rhythms. There are two limitations in existing tools for music error detection: (1) Existing approaches rely on automatic alignment; therefore, they are prone to errors caused by small deviations between alignment targets; (2) There is insufficient data to train music error detection models, resulting in over-reliance on heuristics. To address (1), we propose a novel transformer model, Polytune, that takes audio inputs and outputs annotated music scores. This model can be trained end-to-end to implicitly align and compare performance audio with music scores through latent space representations. To address (2), we present a novel data generation technique capable of creating large-scale synthetic music error datasets. Our approach achieves a 64.1% average Error Detection F1 score, improving upon prior work by 40 percentage points across 14 instruments. Additionally, our model can handle multiple instruments compared with existing transcription methods repurposed for music error detection.

While Signed Distance Fields (SDF) are well-established for modeling watertight surfaces, Unsigned Distance Fields (UDF) broaden the scope to include open surfaces and models with complex inner structures. Despite their flexibility, UDFs encounter significant challenges in high-fidelity 3D reconstruction, such as non-differentiability at the zero level set, difficulty in achieving the exact zero value, numerous local minima, vanishing gradients, and oscillating gradient directions near the zero level set. To address these challenges, we propose Details Enhanced UDF (DEUDF) learning that integrates normal alignment and the SIREN network for capturing fine geometric details, adaptively weighted Eikonal constraints to address vanishing gradients near the target surface, unconditioned MLP-based UDF representation to relax non-negativity constraints, and DCUDF for extracting the local minimal average distance surface. These strategies collectively stabilize the learning process from unoriented point clouds and enhance the accuracy of UDFs. Our computational results demonstrate that DEUDF outperforms existing UDF learning methods in both accuracy and the quality of reconstructed surfaces.

In the coded aperture snapshot spectral imaging system, Deep Unfolding Networks (DUNs) have made impressive progress in recovering 3D hyperspectral images (HSIs) from a single 2D measurement. However, the inherent nonlinear and ill-posed characteristics of HSI reconstruction still pose challenges to existing methods in terms of accuracy and stability. To address this issue, we propose a Mamba-inspired Joint Unfolding Network (MiJUN), which integrates physics-embedded DUNs with learning-based HSI imaging. Firstly, leveraging the concept of trapezoid discretization to expand the representation space of unfolding networks, we introduce an accelerated unfolding network scheme. This approach can be interpreted as a generalized accelerated half-quadratic splitting with a second-order differential equation, which reduces the reliance on initial optimization stages and addresses challenges related to long-range interactions. Crucially, within the Mamba framework, we restructure the Mamba-inspired global-to-local attention mechanism by incorporating a selective state space model and an attention mechanism. This effectively reinterprets Mamba as a variant of the Transformer architecture, improving its adaptability and efficiency. Furthermore, we refine the scanning strategy with Mamba by integrating the tensor mode-k unfolding into the Mamba network. This approach emphasizes the low-rank properties of tensors along various modes, while conveniently facilitating 12 scanning directions. Numerical and visual comparisons on both simulation and real datasets demonstrate the superiority of our proposed MiJUN, and achieving overwhelming detail representation.