Keisuke Maeda

FORUM: Frozen Outputs Reconciled Using Model Agreement for Visual Grounding

Taiyo Sato, Takamasa Sanda, Keisuke Maeda +3

Frozen multimodal large language models (MLLMs) now solve standard referring expression comprehension with a single prompted call, yet on adversarial benchmarks with same-category…

Physics-Guided Flow-Map Matching for Precipitation Nowcasting

Shunya Nagashima, Takumi Bannai, Makoto Misaizu +3

Precipitation nowcasting, generating future radar fields from past observations, is critical for flood warning and disaster response. It is also a demanding benchmark for spatiotem…

Hierarchical Federated Learning with Dynamic Clustering and Adaptive Regularization for Robust Infrastructure Inspection

Yuhu Feng, Keisuke Maeda, Takahiro Ogawa +1

The deployment of data-driven computer vision models for structural health monitoring (SHM) is heavily constrained by the data silo dilemma due to stringent privacy and security re…

Foreground-Aware Dataset Distillation via Dynamic Patch Selection

Longzhen Li, Guang Li, Ren Togo +3

In this paper, we propose a foreground-aware dataset distillation method that enhances patch selection in a content-adaptive manner. With the rising computational cost of training…

Decoupled Audio-Visual Dataset Distillation

Wenyuan Li, Guang Li, Keisuke Maeda +2

Audio-Visual Dataset Distillation aims to compress large-scale datasets into compact subsets while preserving the performance of the original data. However, conventional Distributi…

Objectness Similarity: Capturing Object-Level Fidelity in 3D Scene Evaluation

Yuiko Uchida, Ren Togo, Keisuke Maeda +2

This paper presents Objectness SIMilarity (OSIM), a novel evaluation metric for 3D scenes that explicitly focuses on "objects," which are fundamental units of human visual percepti…

Personalized Federated Learning for Egocentric Video Gaze Estimation with Comprehensive Parameter Frezzing

Yuhu Feng, Keisuke Maeda, Takahiro Ogawa +1

Egocentric video gaze estimation requires models to capture individual gaze patterns while adapting to diverse user data. Our approach leverages a transformer-based architecture, i…

Triplet Synthesis For Enhancing Composed Image Retrieval via Counterfactual Image Generation

Kenta Uesugi, Naoki Saito, Keisuke Maeda +2

Composed Image Retrieval (CIR) provides an effective way to manage and access large-scale visual data. Construction of the CIR model utilizes triplets that consist of a reference i…

Generative Dataset Distillation Based on Self-knowledge Distillation

Longzhen Li, Guang Li, Ren Togo +3

Dataset distillation is an effective technique for reducing the cost and complexity of model training while maintaining performance by compressing large datasets into smaller, more…

Cross-domain Multi-step Thinking: Zero-shot Fine-grained Traffic Sign Recognition in the Wild

Yaozong Gan, Guang Li, Ren Togo +3

In this study, we propose Cross-domain Multi-step Thinking (CdMT) to improve zero-shot fine-grained traffic sign recognition (TSR) performance in the wild. Zero-shot fine-grained T…

Cross-domain Few-shot In-context Learning for Enhancing Traffic Sign Recognition

Yaozong Gan, Guang Li, Ren Togo +3

Recent multimodal large language models (MLLM) such as GPT-4o and GPT-4v have shown great potential in autonomous driving. In this paper, we propose a cross-domain few-shot in-cont…

Reinforcing Pre-trained Models Using Counterfactual Images

Xiang Li, Ren Togo, Keisuke Maeda +2

This paper proposes a novel framework to reinforce classification models using language-guided generated counterfactual images. Deep learning classification models are often traine…

Generative Dataset Distillation: Balancing Global Structure and Local Details

Longzhen Li, Guang Li, Ren Togo +3

In this paper, we propose a new dataset distillation method that considers balancing global structure and local details when distilling the information from a large dataset into a…