Search Results for author: Toshihiko Yamasaki

Found 58 papers, 29 papers with code

Language-Guided Self-Supervised Video Summarization Using Text Semantic Matching Considering the Diversity of the Video

no code implementations14 May 2024 Tomoya Sugihara, Shuntaro Masuda, Ling Xiao, Toshihiko Yamasaki

To address these issues, we analyzed the feasibility in transforming the video summarization into a text summary task and leverage Large Language Models (LLMs) to boost video summarization.

Supervised Video Summarization

Face2Diffusion for Fast and Editable Face Personalization

1 code implementation8 Mar 2024 Kaede Shiohara, Toshihiko Yamasaki

However, it is still challenging for previous methods to preserve both the identity similarity and editability due to overfitting to training samples.

Diffusion Personalization Text-to-Image Generation

Theoretical Understanding of Learning from Adversarial Perturbations

1 code implementation16 Feb 2024 Soichiro Kumano, Hiroshi Kera, Toshihiko Yamasaki

In this study, we provide a theoretical framework for understanding learning from perturbations using a one-hidden-layer network trained on mutually orthogonal samples.

Improving Plasticity in Online Continual Learning via Collaborative Learning

1 code implementation1 Dec 2023 Maorong Wang, Nicolas Michel, Ling Xiao, Toshihiko Yamasaki

To this end, we propose Collaborative Continual Learning (CCL), a collaborative learning based strategy to improve the model's capability in acquiring new concepts.

Continual Learning

Rethinking Momentum Knowledge Distillation in Online Continual Learning

no code implementations6 Sep 2023 Nicolas Michel, Maorong Wang, Ling Xiao, Toshihiko Yamasaki

While Knowledge Distillation (KD) has been extensively used in offline Continual Learning, it remains under-exploited in OCL, despite its potential.

Continual Learning Knowledge Distillation

New metrics for analyzing continual learners

no code implementations1 Sep 2023 Nicolas Michel, Giovanni Chierchia, Romain Negrel, Jean-François Bercher, Toshihiko Yamasaki

This scenario, known as Continual Learning (CL) poses challenges to standard learning algorithms which struggle to maintain knowledge of old tasks while learning new ones.

Continual Learning

BSED: Baseline Shapley-Based Explainable Detector

no code implementations14 Aug 2023 Michihiro Kuroki, Toshihiko Yamasaki

The Shapley value can attribute the prediction of a learned model to a baseline feature while satisfying the explainability axioms.

Attribute Explainable artificial intelligence +4

Personalized Image Enhancement Featuring Masked Style Modeling

1 code implementation15 Jun 2023 Satoshi Kosugi, Toshihiko Yamasaki

Second, to allow this model to consider the contents of images, we propose a novel training scheme where we download images from Flickr and create pseudo input and retouched image pairs using a degrading model.

Image Enhancement Language Modelling +1

Crowd-Powered Photo Enhancement Featuring an Active Learning Based Local Filter

1 code implementation15 Jun 2023 Satoshi Kosugi, Toshihiko Yamasaki

Existing photo enhancement methods are either not content-aware or not local; therefore, we propose a crowd-powered local enhancement method for content-aware local enhancement, which is achieved by asking crowd workers to locally optimize parameters for image editing functions.

Active Learning

Online Open-set Semi-supervised Object Detection with Dual Competing Head

no code implementations23 May 2023 Zerun Wang, Ling Xiao, Liuyu Xiang, Zhaotian Weng, Toshihiko Yamasaki

To alleviate these issues, this paper proposes an end-to-end online OSSOD framework that improves performance and efficiency: 1) We propose a semi-supervised outlier filtering method that more effectively filters the OOD instances using both labeled and unlabeled data.

object-detection Object Detection +1

Toward Extremely Lightweight Distracted Driver Recognition With Distillation-Based Neural Architecture Search and Knowledge Transfer

1 code implementation9 Feb 2023 Dichao Liu, Toshihiko Yamasaki, Yu Wang, Kenji Mase, Jien Kato

Experimental results on the Statefarm Distracted Driver Detection Dataset and AUC Distracted Driver Dataset show that the proposed approach is highly effective for recognizing distracted driving behaviors from photos: (1) the teacher network's accuracy surpasses the previous best accuracy; (2) the student network achieves very high accuracy with only 0. 42M parameters (around 55% of the previous most lightweight model).

Knowledge Distillation Neural Architecture Search +1

Attribute-Guided Multi-Level Attention Network for Fine-Grained Fashion Retrieval

1 code implementation27 Dec 2022 Ling Xiao, Toshihiko Yamasaki

Our model consistently outperforms existing attention based methods when assessed on the FashionAI (62. 8788% in MAP), DeepFashion (8. 9804% in MAP), and Zappos50k datasets (93. 32% in Prediction accuracy).

Attribute Image Classification +3

Semi-supervised Fashion Compatibility Prediction by Color Distortion Prediction

no code implementations27 Dec 2022 Ling Xiao, Toshihiko Yamasaki

In this paper, we propose a general color distortion prediction task forcing the baseline to recognize low-level image information to learn more discriminative representation for fashion compatibility prediction.

Semi-Supervised Fashion Compatibility

Fine-Grained Image Style Transfer with Visual Transformers

1 code implementation11 Oct 2022 Jianbo Wang, Huan Yang, Jianlong Fu, Toshihiko Yamasaki, Baining Guo

Such a design usually destroys the spatial information of the input images and fails to transfer fine-grained style patterns into style transfer results.

Style Transfer

Improving Robustness to Out-of-Distribution Data by Frequency-based Augmentation

no code implementations6 Sep 2022 Koki Mukai, Soichiro Kumano, Toshihiko Yamasaki

In order to improve the robustness against out-of-distribution data, we present a frequency-based data augmentation technique that replaces the frequency components with other images of the same class.

Data Augmentation

SAT: Self-adaptive training for fashion compatibility prediction

1 code implementation25 Jun 2022 Ling Xiao, Toshihiko Yamasaki

Then, we propose a self-adaptive triplet loss (SATL), where the DS of the outfit is considered.

Superclass Adversarial Attack

no code implementations29 May 2022 Soichiro Kumano, Hiroshi Kera, Toshihiko Yamasaki

Adversarial attacks have only focused on changing the predictions of the classifier, but their danger greatly depends on how the class is mistaken.

Adversarial Attack Multi-Label Classification

Green Hierarchical Vision Transformer for Masked Image Modeling

1 code implementation26 May 2022 Lang Huang, Shan You, Mingkai Zheng, Fei Wang, Chen Qian, Toshihiko Yamasaki

We present an efficient approach for Masked Image Modeling (MIM) with hierarchical Vision Transformers (ViTs), allowing the hierarchical ViTs to discard masked patches and operate only on the visible ones.

Object Detection

Detecting Deepfakes with Self-Blended Images

1 code implementation CVPR 2022 Kaede Shiohara, Toshihiko Yamasaki

In this paper, we present novel synthetic training data called self-blended images (SBIs) to detect deepfakes.

DeepFake Detection

Learning Where to Learn in Cross-View Self-Supervised Learning

1 code implementation CVPR 2022 Lang Huang, Shan You, Mingkai Zheng, Fei Wang, Chen Qian, Toshihiko Yamasaki

In this paper, we present a new approach, Learning Where to Learn (LEWEL), to adaptively aggregate spatial information of features, so that the projected embeddings could be exactly aligned and thus guide the feature learning better.

object-detection Object Detection +3

Edge-Level Explanations for Graph Neural Networks by Extending Explainability Methods for Convolutional Neural Networks

no code implementations1 Nov 2021 Tetsu Kasanishi, Xueting Wang, Toshihiko Yamasaki

Graph Neural Networks (GNNs) are deep learning models that take graph data as inputs, and they are applied to various tasks such as traffic prediction and molecular property prediction.

Molecular Property Prediction Property Prediction +1

Learning from Synthetic Shadows for Shadow Detection and Removal

1 code implementation5 Jan 2021 Naoto Inoue, Toshihiko Yamasaki

To overcome this challenge, we present SynShadow, a novel large-scale synthetic shadow/shadow-free/matte image triplets dataset and a pipeline to synthesize it.

Shadow Detection And Removal Shadow Removal

Predicting Online Video Advertising Effects with Multimodal Deep Learning

no code implementations22 Dec 2020 Jun Ikeda, Hiroyuki Seshime, Xueting Wang, Toshihiko Yamasaki

With expansion of the video advertising market, research to predict the effects of video advertising is getting more attention.

Multimodal Deep Learning

Are DNNs fooled by extremely unrecognizable images?

1 code implementation7 Dec 2020 Soichiro Kumano, Hiroshi Kera, Toshihiko Yamasaki

In this paper, we address the question of whether there can be fooling images with no characteristic pattern of natural objects locally or globally.

Out-of-Distribution Detection

Image inpainting using frequency domain priors

1 code implementation3 Dec 2020 Hiya Roy, Subhajit Chaudhury, Toshihiko Yamasaki, Tatsuaki Hashimoto

To alleviate these problems, we investigate if it is possible to obtain better performance by training the networks using frequency domain information (Discrete Fourier Transform) along with the spatial domain information.

Image Inpainting

Re-identification = Retrieval + Verification: Back to Essence and Forward with a New Metric

1 code implementation23 Nov 2020 Zheng Wang, Xin Yuan, Toshihiko Yamasaki, Yutian Lin, Xin Xu, Wenjun Zeng

In essence, current re-ID overemphasizes the importance of retrieval but underemphasizes that of verification, \textit{i. e.}, all returned images are considered as the target.

Image Retrieval Retrieval

Pretext-Contrastive Learning: Toward Good Practices in Self-supervised Video Representation Leaning

1 code implementation29 Oct 2020 Li Tao, Xueting Wang, Toshihiko Yamasaki

It is convenient to treat PCL as a standard training strategy and apply it to many other works in self-supervised video feature learning.

Contrastive Learning Data Augmentation +4

Self-supervised Video Representation Learning Using Inter-intra Contrastive Framework

2 code implementations6 Aug 2020 Li Tao, Xueting Wang, Toshihiko Yamasaki

With the proposed Inter-Intra Contrastive (IIC) framework, we can train spatio-temporal convolutional networks to learn video representations.

Action Recognition In Videos Contrastive Learning +6

Image Aesthetics Prediction Using Multiple Patches Preserving the Original Aspect Ratio of Contents

no code implementations5 Jul 2020 Lijie Wang, Xueting Wang, Toshihiko Yamasaki

The spread of social networking services has created an increasing demand for selecting, editing, and generating impressive images.

Motion Representation Using Residual Frames with 3D CNN

3 code implementations21 Jun 2020 Li Tao, Xueting Wang, Toshihiko Yamasaki

In this paper, we propose a fast but effective way to extract motion features from videos utilizing residual frames as the input data in 3D ConvNets.

Action Recognition Optical Flow Estimation

Investigating Generalization in Neural Networks under Optimally Evolved Training Perturbations

1 code implementation14 Mar 2020 Subhajit Chaudhury, Toshihiko Yamasaki

In this paper, we study the generalization properties of neural networks under input perturbations and show that minimal training data corruption by a few pixel modifications can cause drastic overfitting.

Domain Adaptation

Rethinking Motion Representation: Residual Frames with 3D ConvNets for Better Action Recognition

3 code implementations16 Jan 2020 Li Tao, Xueting Wang, Toshihiko Yamasaki

Further analysis indicates that better motion features can be extracted using residual frames with 3D ConvNets, and our residual-frame-input path is a good supplement for existing RGB-frame-input models.

Action Recognition Optical Flow Estimation

Assessing Robustness of Deep learning Methods in Dermatological Workflow

no code implementations15 Jan 2020 Sourav Mishra, Subhajit Chaudhury, Hideaki Imaizumi, Toshihiko Yamasaki

This paper aims to evaluate the suitability of current deep learning methods for clinical workflow especially by focusing on dermatology.

Weakly Supervised Video Summarization by Hierarchical Reinforcement Learning

no code implementations12 Jan 2020 Yiyan Chen, Li Tao, Xueting Wang, Toshihiko Yamasaki

For each subtask, the manager is trained to set a subgoal only by a task-level binary label, which requires much fewer labels than conventional approaches.

Hierarchical Reinforcement Learning reinforcement-learning +2

Unpaired Image Enhancement Featuring Reinforcement-Learning-Controlled Image Editing Software

no code implementations17 Dec 2019 Satoshi Kosugi, Toshihiko Yamasaki

This paper tackles unpaired image enhancement, a task of learning a mapping function which transforms input images into enhanced images in the absence of input-output image pairs.

Image Enhancement reinforcement-learning +1

Improving image classifiers for small datasets by learning rate adaptations

no code implementations26 Mar 2019 Sourav Mishra, Toshihiko Yamasaki, Hideaki Imaizumi

Our paper introduces an efficient combination of established techniques to improve classifier performance, in terms of accuracy and training time.

Fully Convolutional Network with Multi-Step Reinforcement Learning for Image Processing

1 code implementation10 Nov 2018 Ryosuke Furuta, Naoto Inoue, Toshihiko Yamasaki

This paper tackles a new problem setting: reinforcement learning with pixel-wise rewards (pixelRL) for image processing.

Image Denoising Image Restoration +3

Fast and Robust Estimation for Unit-Norm Constrained Linear Fitting Problems

no code implementations CVPR 2018 Daiki Ikami, Toshihiko Yamasaki, Kiyoharu Aizawa

M-estimator using iteratively reweighted least squares (IRLS) is one of the best-known methods for robust estimation.

Local and Global Optimization Techniques in Graph-Based Clustering

no code implementations CVPR 2018 Daiki Ikami, Toshihiko Yamasaki, Kiyoharu Aizawa

We propose a local optimization method, which is widely applicable to graph-based clustering cost functions.

Clustering

Object Detection for Comics using Manga109 Annotations

5 code implementations23 Mar 2018 Toru Ogawa, Atsushi Otsubo, Rei Narita, Yusuke Matsui, Toshihiko Yamasaki, Kiyoharu Aizawa

We annotated an existing image dataset of comics and created the largest annotation dataset, named Manga109-annotations.

Object object-detection +1

Supervised classification of Dermatological diseases by Deep learning

no code implementations11 Feb 2018 Sourav Mishra, Toshihiko Yamasaki, Hideaki Imaizumi

This paper introduces a deep-learning based efficient classifier for common dermatological conditions, aimed at people without easy access to skin specialists.

Classification General Classification

PQTable: Non-exhaustive Fast Search for Product-quantized Codes using Hash Tables

no code implementations21 Apr 2017 Yusuke Matsui, Toshihiko Yamasaki, Kiyoharu Aizawa

In this paper, we propose a product quantization table (PQTable); a fast search method for product-quantized codes via hash-tables.

Quantization

Uncalibrated Photometric Stereo by Stepwise Optimization Using Principal Components of Isotropic BRDFs

no code implementations CVPR 2016 Keisuke Midorikawa, Toshihiko Yamasaki, Kiyoharu Aizawa

We propose a model that represents various isotropic reflectance functions by using the principal components of items in a dataset, and formulate the uncalibrated photometric stereo as a regression problem.

PQTable: Fast Exact Asymmetric Distance Neighbor Search for Product Quantization Using Hash Tables

no code implementations ICCV 2015 Yusuke Matsui, Toshihiko Yamasaki, Kiyoharu Aizawa

We propose the product quantization table (PQTable), a product quantization-based hash table that is fast and requires neither parameter tuning nor training steps.

Quantization

Sketch-based Manga Retrieval using Manga109 Dataset

no code implementations15 Oct 2015 Yusuke Matsui, Kota Ito, Yuji Aramaki, Toshihiko Yamasaki, Kiyoharu Aizawa

From the experiments, we verified that: (1) the retrieval accuracy of the proposed method is higher than those of previous methods; (2) the proposed method can localize an object instance with reasonable runtime and accuracy; and (3) sketch querying is useful for manga search.

Quantization Retrieval +1

Cannot find the paper you are looking for? You can Submit a new open access paper.