Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Masao SakauchiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works508 from public data
- The SR-tree1,050
This paper proposes a new index structure called the SR-tree (Sphere/Rectangle-tree) which integrates bounding spheres and bounding rectangles which enhances the performance on nearest neighbor queries especially for high-dimensional and non-uniform data which can be practical in actual image/video similarity indexing.
- Embedding Watermarks into Deep Neural Networks812
This work proposes to use digital watermarking technology to protect intellectual property and detect intellectual property infringement in the use of trained models, and proposes a general framework for embedding a watermark in model parameters, using a parameter regularizer.
- 556
- Name-It: naming and detecting faces in news videos316
Name-It, a system that associates faces and names in news videos, takes a multimodal video analysis approach: face sequence extraction and similarity evaluation from videos, name extraction from transcripts, and video-caption recognition.
- 313
- Joint Detection and Recounting of Abnormal Events by Learning Deep Generic Knowledge278
This paper addresses the problem of joint detection and recounting of abnormal events in videos by integrating a generic CNN model and environment-dependent anomaly detectors and produces promising results of abnormal event recounting.
- Video OCR: indexing digital news libraries by recognition of superimposed captions211
To solve two problems of character recognition for videos, low-resolution characters and extremely complex backgrounds, an interpolation filter, multi-frame integration and character extraction filters are applied and the overall recognition results are satisfactory for use in news indexing.
- MADGAN: unsupervised medical anomaly detection GAN using multiple adjacent brain MRI slice reconstruction210
This work proposes unsupervised medical anomaly detection generative adversarial network (MADGAN), a novel two-step method using GAN-based multiple adjacent brain MRI slice reconstruction to detect brain anomalies at different stages on multi-sequence structural MRI.
- Physical Adversarial Attack Meets Computer Vision: A Decade Survey169
A comprehensive overview of physical adversarial attacks is presented and a proposed evaluation metric, hiPAA, comprises six perspectives: Effectiveness, Stealthiness, Robustness, Practicability, Aesthetics, and Economics.
- 161
- 155
- Faster R-CNN Features for Instance Search153
This work explores the suitability for instance retrieval of image-and region-wise representations pooled from an object detection CNN such as Faster R-CNN and takes advantage of the object proposals learned by a Region Proposal Network (RPN) and their associated CNN features to build an instance search pipeline.
- Guidance and Evaluation: Semantic-Aware Image Inpainting for Mixed Scenes139
This paper proposes a Semantic Guidance and Evaluation Network (SGE-Net) to iteratively update the structural priors and the inpainted image in an interplay framework of semantics extraction and image inpainting, which utilizes semantic segmentation map as guidance in each scale of inPainting.
- Cascaded SR-GAN for Scale-Adaptive Low Resolution Person Re-identification123
A unified formulation, a Cascaded Super-Resolution GAN (CSR-GAN) framework is proposed, which requires a network not only synthesizing high-resolution images with different upscaling factors, but also extracting discriminative image feature for judging person’s identity.
- Image Inpainting Guided by Coherence Priors of Semantics and Textures118
A multi-scale joint optimization framework is adopted to first model the coherence priors and then accordingly interleaving optimize image inpainting and semantic segmentation in a coarse-to-fine manner and two coherence losses are proposed to constrain the consistency between the semantics and the inpainted image in terms of the overall structure and detailed textures.
- Cooking navi117
According to the result of a preliminary experiment, all users from novice to experienced cooks could finish two dishes in parallel while enjoyeing the cooking very much, shows the effectiveness of the multimedia navigation that is proposed.
- 109
- Learning More with Less102
This paper proposes Conditional Progressive Growing of GANs (CPGGANs), incorporating highly-rough bounding box conditions incrementally into PGGANs to place brain metastases at desired positions/sizes on 256 × 256 Magnetic Resonance (MR) images, for Convolutional Neural Network-based tumor detection; this first GAN-based medical DA using automatic boundingbox annotation improves the training robustness.
- Learning More with Less: Conditional PGGAN-based Data Augmentation for Brain Metastases Detection Using Highly-Rough Annotation on MR Images102
This paper proposes Conditional Progressive Growing of GANs (CPGGANs), incorporating highly-rough bounding box conditions incrementally into PGGANs to place brain metastases at desired positions/sizes on 256 × 256 Magnetic Resonance (MR) images, for Convolutional Neural Network-based tumor detection; this first GAN-based medical DA using automatic boundingbox annotation improves the training robustness.
- Building Compact Local Pairwise Codebook with Joint Feature Space Clustering100
Experimental results on challenging datasets demonstrate that LPC outperforms the baselines and performs competitively against the state-of-the-art techniques in scene and object categorization tasks where a large number of categories need to be recognized.
- 99
- 89
- 88
- Evaluation of deep convolutional neural networks for glaucoma detection87
DCNNs may be a useful tool for detecting glaucoma or glauca-suspected eyes by use of fundus color images and proper preprocessing and collection of qualified images are essential to improving the discriminative ability.
- Learning Sparse and Identity-Preserved Hidden Attributes for Person Re-Identification82
This work proposes to incrementally generate Deep Hidden Attribute (DHA) based on baseline deep network for newly uncovered annotations based on an auto-encoder model that can be plugged into any deep network to mine latent information in an unsupervised manner.
- 77
- 73
- 71
- Beyond Intra-modality: A Survey of Heterogeneous Person Re-identification69
This paper provides a comprehensive review of state-of-the-art Hetero-ReID methods that address the challenge of inter-modality discrepancies and compares the representative approaches from two perspectives, i.e., the application scenario and the learning pipeline.
- Alleviating Cold-Start Problems in Recommendation through Pseudo-Labelling over Knowledge Graph61
This study proposes a knowledge graph (KG)-aware recommender based on graph neural networks, which augments labelled samples through pseudo-labelling and improves recommendation performance for cold-start users/items.
- 60
- Geo-Localization via Ground-to-Satellite Cross-View Image Retrieval58
This paper investigates an extremely challenging task: given a ground-view image of a landmark, it aims to achieve cross-view geo-localization by searching out its corresponding satellite-view images, and proposes a Peer Learning and Cross Diffusion (PLCD) framework.
- Efficient Image Retrieval via Decoupling Diffusion into Online and Offline Processing58
This work precomputes the diffusion results of each element in the database, making the online search a simple linear combination on top of the k-NN search process, and becomes 10∼ times faster in terms of online search speed.
- Topic Threading for Structuring a Large-Scale News Video Archive58
A structuring method applied to the accumulated news videos is introduced and a user interface for topic thread-based news video retrieval is introduced to lessen the tedious selection during a tracking process by a user.
- 57
- Towards Unsupervised Crowd Counting via Regression-Detection Bi-knowledge Transfer52
This paper introduces an iterative self-supervised learning scheme with regression-detection bi-knowledge transfer in the target and formulate the mutual transformations between the outputs of regression- and detection-based models as two scene-agnostic transformers which enable knowledge distillation between the two models.
- Large vocabulary quantization for searching instances from videos52
This paper proposed an algorithm for instance search that outperformed all submissions on the instance search dataset TRECVID 2011, and showed that the top performance is mainly due to similar scene retrieval, instead of the same instance search.
- NII-HITACHI-UIT at TRECVID 2016.49
The proposed system for TRECVID In-stance Search Task with novel compound query type achieved very high recognition rate and is ranked at the third place for fully automatic run.
- 47
- DoT-GNN45
The key to the success of DoT-GNN is that the transferred graph addresses the challenge of the appearance change, while the graph representation in GNN overcomes the challengeof the layout and membership change.
- Name-It: Naming and Detecting Faces in Video by the Integration of Image and Natural Language Processing45
The proposed Name-It system, a system that associates faces and names in news videos, takes full advantage of advanced image and natural language processing and effectively extracts names by using lexical/grammatical analysis and knowledge of the news video topics structure.
- 42
- 42
- 40
- 40
- Gastric Cancer Detection from Endoscopic Images Using Synthesis by GAN39
This paper proposes a method that lessens this dataset bias by generating new images using a generative model, which synthesizes an image from two images in a dataset.
- 39
- Temporal Matching Kernel with Explicit Feature Maps38
A novel query expansion method that takes advantage of the joint descriptor/timestamp representation to automatically align the first result set and produce an enriched temporal query, and improves the localization trade-off between quality and representation size by using several complementary temporal match kernels.
- 38
- Unsupervised Foggy Scene Understanding via Self Spatial-Temporal Label Diffusion37
A novel Target-Domain driven pseudo label Diffusion (TDo-Dif) scheme that employs superpixels and optical flows to identify the spatial similarity and temporal correspondence of the sequential image data, and diffuses the confident but sparse pseudo labels within a superpixel or a temporal corresponding pair linked by the flow.
- 37
- 37
- 36
- Evaluation of multiple features for violent scenes detection36
This study rigorously analyzed a set of low-level features and a deep learning feature that captures the appearance, color, texture, motion and audio in video and provides an empirical foundation for selecting feature sets that are capable of dealing with heterogeneous content comprising violent scenes in movies.
- Mining on Heterogeneous Manifolds for Zero-Shot Cross-Modal Image Retrieval35
This work proposes a bi-directional random walk scheme to mining more reliable relationships between images by traversing heterogeneous manifolds in the feature space of each modality to alleviate the interference caused by noisy similarities in the cross-modal feature space.
- Instance search retrospective with focus on TRECVID35
An overview of the Video instance search benchmark which was run over a period of 6 years (2010–2015) as part of the TREC Video Retrieval workshop series is presented.
- 35
- 34
- Improving Camouflaged Object Detection with the Uncertainty of Pseudo-edge Labels33
This paper proposes a new framework that makes full use of multiple visual cues, i.e., saliency as well as edges, to refine the predicted camouflaged map, and proposes an uncertainty-based module to reduce the uncertainty and noise of such two pseudo labels.
- When Pedestrian Detection Meets Nighttime Surveillance: A New Benchmark32
This paper builds a novel pedestrian detection dataset from the nighttime surveillance aspect, and compares the performances of state-of-the-art pedestrian detectors and reveals that the methods cannot solve all the challenging problems of NightSurveillance.
- 32
- Human gesture recognition system for TV viewing using time-of-flight camera32
A new device-free user interface for TV viewing that uses a human gesture recognition technique that recognizes a large enough variety of gestures and is useful because it does not require any contact-type devices.
- Simple low-dimensional features approximating NCC-based image matching32
New low-dimensional image features that enable images to be very efficiently matched and are shown to outperform other methods: multilevel successive elimination algorithm (MSEA), discrete cosine transform (DCT) coefficients, and histograms, achieving very high precision while only slightly sacrificing recall.
- NII-UIT at MediaEval 2015 Affective Impact of Movies Task31
Experimental results show that the unied framework can detect violent videos and its aective impact with a reasonable accuracy and using deep features can signicantly improve the detection performance of both subtasks.
- 30
- Learning More with Less: GAN-based Medical Image Augmentation29
This paper introduces GAN-based Medical Image Augmentation, along with tricks to boost classification/object detection/segmentation performance using them, based on the experience and related work.
- Vasopressin mediated vasodilation of cerebral arteries29
The results suggest that the arteries of the circle of Willis at the base of the brain are more sensitive to nitric oxide release induced by vasopressin compared with other intracranial and extracranial arteries.
- 28
- An adjustable-purpose image watermarking technique by particle swarm optimization28
This paper proposes an adjustable-purpose, reversible and fragile watermarking scheme for imageWatermarking by particle swarm optimization (PSO), and shows that the proposed technique has accomplished higher capacity and higher PSNR (peak signal-to-noise ratio) water marking.
- Structured learning of local features for human action classification and localization28
This work introduces a novel feature selection process called Sparse Hierarchical Bayes Filter and presents the application of structured learning in human action analysis, and proposes and shows that, by taking into account global configuration of local features, it can greatly improve recognition performance.
- Improving Adversarial Robustness via Information Bottleneck Distillation27
This study closely examines the information bottleneck principle and proposes an Information Bottleneck Dis-tillation approach that utilizes prior knowledge obtained from a robust pre-trained model to boost information bottlenecks.
- 27
- Region-Based Image Retrieval Revisited27
This work revisits RBIR by incorporating semantic specification of objects and intuitive specification of spatial relationships, and proposes a multitask CNN feature that allows us to use deep learning technique and to jointly handle multi-aspect object specification.
- 27
- A FRAMEWORK FOR VIDEO SEGMENTATION USING GLOBAL AND LOCAL FEATURES27
A framework for scene detection mainly based on entropy and Speeded Up Robust Features (SURF) features is proposed, which detects the fade and abrupt boundaries based on frame entropy analysis and SURF features matching.
- Scene duplicate detection from videos based on trajectories of feature points27
A method to detect scene duplicates, another type of near duplicate that is composed of different footages taking the same scene, the same event, at the same time, but from the different viewpoints, is proposed.
- National institute of informatics, japan at TRECVID 2007: BBC rushes summarization26
In this paper, a method for BBC rushes summarization is presented, first the input video is decomposed into fragments by comparing consecutive frames, then these fragments are grouped by a clustering method, and consecutive segments are grouped into segments.
- Mining Large-Scale Broadcast Video Archives Towards Inter-video Structuring26
This paper designs and constructed a broadcast video archive system having sufficient capacity and functionality to serve as the testbed for indexing and mining research, and introduces research efforts utilizing the archives as huge multimedia corpora.
- 25
- 25
- 25
- 24
- Learning to Attack Real-World Models for Person Re-identification via Virtual-Guided Meta-Learning24
This study argues that learning powerful attackers with high universality that works well on unseen domains is an important step in promoting the robustness of re-ID systems and introduces a novel universal attack algorithm called ``MetaAttack'' for person re- ID.
- Classification of large-scale image database of various skin diseases using deep learning24
The experimental results demonstrated that the proposed classification reinforced by metric learning and aggregation of multiple input images was effective in the classification of patients with diverse skin diseases and imaging conditions.
- Discriminative Learning of Open-Vocabulary Object Retrieval and Localization by Negative Phrase Augmentation24
This paper proposes Query-Adaptive R-CNN, a simple extension of Faster R- CNN adapted to open-vocabulary queries, and proposes negative phrase augmentation (NPA) to mine hard negative samples which are visually similar to the query and at the same time semantically mutually exclusive of the query.
- Incremental Deep Hidden Attribute Learning24
This work proposes a new person representation with hidden attributes by mining latent information from visual feature in an unsupervised manner using an auto-encoder model plugged-in to the deep learning network to compose a Deep Hidden Attribute Network (DHA-Net).
- Multilevel Thresholding Color Image Segmentation Using a Modified Artificial Bee Colony Algorithm23
In order to improve the local search ability of ABC algorithm, Krill Herd algorithm is incorporated into its onlooker bees phase and the proposed algorithm is named as Krill herd-inspired modified Artificial Bee Colony algorithm (KABC algorithm).
- Reconfigurable Inverted Index23
A reconfigurable inverted index (Rii) is developed that achieves a comparable performance with state-of-the art systems such as Faiss.
- Learning Directional Local Pairwise Bases with Sparse Coding23
Directional Local Pairwise Bases (DLPB) is proposed that applies sparse coding to learn a compact set of bases capturing correlation between these descriptors, so to avoid the combinatorial explosion.
- 23
- Beyond Domain Gap: Exploiting Subjectivity in Sketch-Based Person Retrieval22
A non-local (NL) fusion module that gathers sketches from different witnesses for the same identity is proposed and an AttrAlign module utilizes attributes as an implicit mask to align cross-domain features.
- 22
- 22
- Scene duplicate detection based on the pattern of discontinuities in feature point trajectories22
An acceleration method based on the discontinuity pattern, which is more robust to variations in camerawork and editing than conventional features, to dramatically reduce the computation burden is developed.
- 22
- 21
- BIG-OH: BInarization of gradient orientation histograms21
BIG-OH, a simple yet extremely effective method for binary quantization of any descriptor based on gradient orientation histograms, is proposed and has performance comparable to or better than other descriptors, including BRISK, CARD, BRIEF, D-BriEF, SQ, and PCA-SIFT.
- 21
- An efficient method for face retrieval from large video datasets21
In order to make the face retrieval robust, the faces of the same person appearing in individual shots are grouped into a single face track by using a reliable tracking method, which can achieve high accuracy in retrieval while maintaining low computational cost.
- 20
- Multiple Object Tracking from appearance by hierarchically clustering tracklets20
This work explores MOT using object appearances as the main source of association between objects in a video, using spatial and temporal priors as weighting factors, and form initial tracklets by leveraging on the idea that instances of an object that are close in time should be similar in appearance.
- GAN-Based Multiple Adjacent Brain MRI Slice Reconstruction for Unsupervised Alzheimer’s Disease Diagnosis20
This work proposes a two-step method using Generative Adversarial Network-based multiple adjacent brain MRI slice reconstruction to detect Alzheimer's Disease at various stages and shows that it can reliably detectAD at a very early stage with Area Under the Curve (AUC) 0.780 while also detecting AD at a late stage much more accurately with AUC 0.917.
- 20
- 19
- NII-UIT at MediaEval 2014 violent scenes detection affect task19
This paper describes the VSD system used in MediaEval 2014 and discusses the performance results obtained in main subjective tasks, focusing on improving the trajectory-based motion features that have been proven eective in previous year’s evaluation.
- 19
- A Novel Approach for Filtering Junk Images from Google Search Results19
A kernel-based image clustering technique to partition the images returned by Google Images into multiple visually-similar clusters is developed and a novel framework is developed to achieve more intuitive visualization of large amount of returned images according to their visual similarity.
- Using Topic Concepts for Semantic Video Shots Classification19
This paper proposes to enhance the classification of high-level concepts using intermediate topic concepts and study various fusion strategies to combine topic concepts with visual features in order to outperform unimodal classifiers.
- Matting by Generation18
An innovative approach for image matting that redefines the traditional regression-based task as a generative modeling challenge and harnesses the capabilities of latent diffusion models, enriched with extensive pre-trained knowledge, to regularize the matting process.
- Imageability- and Length-Controllable Image Captioning18
This paper proposes an image captioning method which can generate both imageability- and length-controllable captions, and shows that the proposed method provides a promising step towards tailoring image captions closer to certain applications.
- 18
- Concept-Level Multimodal Ranking of Flickr Photo Tags via Recall Based Weighting18
This work proposes a tag ranking scheme based on voting from photo neighbors derived from multimodal information that determines photo neighbors leveraging geo, visual, and semantics concepts derived from spatial information, visual content, and textual metadata, respectively.
- 18
- 18
- 18
- Reference-Guided Texture and Structure Inference for Image Inpainting17
Both quantitative and qualitative evaluations demonstrate the superiority of the method over the state-of-the-art methods in terms of completing complex holes.
- Neural Global Shutter: Learn to Restore Video from a Rolling Shutter Camera with Global Reset Feature17
This paper investigates using rolling shutter with a global reset feature (RSGR) to restore clean global shutter (GS) videos and develops a novel algorithm incorporating spatial and temporal designs to correct the spatial-varying RSGR distortion.
- Poses Guide Spatiotemporal Model for Vehicle Re-identification17
A vehicles’ Poses Guide Spatiotemporal model (PGST) for assisting vehicle Re-ID is proposed which fuses the results of visual appearance model and spatiotem temporal model together and achieves competitive performances.
- 17
- 17
- 16
- When face-tracking meets social networks: a story of politics in news videos16
A system deriving multiple social networks from face detections in news videos is introduced, which is used as a support for political studies and an overview of the political landscape presented by the Japanese public broadcaster NHK over a decade of the 7 PM news archives is provided.
- Robust Recognition of Specific Human Behaviors in Crowded Surveillance Video Sequences16
A method that can detect specific human behaviors even in crowded surveillance video scenes and precisely identified specific behaviors and achieved first place for detecting running people in the TRECVID 2009 Surveillance Event Detection Task is described.
- trackThem: Exploring a Large-Scale News Video Archive by Tracking Human Relations16
The idea and the implementation of the method are presented, and the trackThem interface that enables the retrieval and at the same time track down the relations are introduced, and interesting relations are found through the exploration of the archive by making use of the proposed interface.
- 15
- VabCut: A video extension of GrabCut for unsupervised video foreground object segmentation15
A new iterative and collaborative method for an optimal frame alignment, based on points of interest and RANSAC, which automatically discards outliers and refines the homographies in turns is introduced, which outperforms the state of the art methods while being faster.
- 15
- 15
- mediaWalker15
A video browsing interface 'mediaWalker' is introduced that lets users explore a news video archive based on a time-series semantic structure; the 'topic thread' structure.
- A News Video Browser Using Identical Video Segment Detection15
This work searched identical video segments from 350 hours-long broadcast news video archive and implemented a news video browser using results of identical video segment detection, and found that time intervals of appearance of the same video segment can be used for classify identicalVideo segments.
- Referring Image Segmentation via Joint Mask Contextual Embedding Learning and Progressive Alignment Network14
JMCELN is designed to enhance the Cascade Framework by incorporating a Learnable Contextual Embedding and a Progressive Alignment Network (PAN), which guides the Learnable Contextual Embedding to incorporate more discriminative information for reasoning, leading to enhanced prediction quality and a reduction in error propagation.
- Face Detection, Tracking, and Classification from Large-Scale News Archives for Analysis of Key Political Figures14
A deep learning-based method combining face detection, tracking, and classification, which is particularly unique because it does not require any re-training when targeting new individuals, is presented.
- Continuous and Gradual Style Changes of Graphic Designs with Generative Model14
A novel deep generative model is proposed that enables the generation of various layout designs and guarantees continuous and gradual changes in layouts, for effectively exploring graphic designs.
- Face Retrieval in Large-Scale News Video Datasets14
Approaches for face retrieval that are scalable to large-scale video datasets while maintaining competitive performances with state-of-the-art approaches are introduced and a remarkable balance between accuracy andiency is achieved.
- Real Time Tunnel Based Video Summarization Using Direct Shift Collision Detection14
The proposed film map generation technique is used to summarize a video which leaves just-in-time renderer to render only necessary frames and reveals an overall performance that gives us real time results without losing the main contents of summarized video.
- 14
- 14
- Multimedia Integration for Cooking Video Indexing14
An integration system that performs semantic segmentations of video and text and associates them together is proposed, and through this method, many applications should become possible, such as a cooking navigation software.
- Scene Text Detection with Polygon Offsetting and Border Augmentation13
This work presents a segmentation-based text detector which can detect an arbitrary shaped scene text by using polygon offsetting, combined with the border augmentation, which better distinguishes contiguous and arbitrary shaped text instances from nearby non-text regions.
- Statistical Inference of Gaussian-Laplace Distribution for Person Verification13
Experiments conducted on a metric learning method by exploiting Gaussian-Laplace distribution statistical inference show its superiority in terms of effectiveness and efficiency as compared with the state-of-the-art approaches, no matter the appearance description is handcrafted or deep learned.
- Multimedia Event Detection Using Event-Driven Multiple Instance Learning13
This work proposes a new method, Event-driven Multiple Instance Learning (EDMIL), to learn the key evidences for event detection, which treats each segment as an instance and quantize the instance-event similarity into different levels of relatedness.
- 13
- Annotation propagation in image databases using similarity graphs13
An influence propagation strategy, SW-KProp, that requires no human intervention beyond the initial labeling of a subset of the images and which enhances the quality of the similarity graph by selecting a reduced feature set for each prelabeled image and rebuilding its neighborhood.
- Integrating local action elements for action analysis13
This paper proposes a framework for human action analysis from video footage that marginally outperforms all existing state-of-the-arts techniques on a range of different datasets and considers two different recipes to construct action elements.
- Indexing local configurations of features for scalable content-based video copy detection13
A computationally inexpensive solution is suggested for including a minimal amount of configuration information that significantly improves the balance between overall detection quality and scalability in content-based video copy detection.
- An Efficient Feature Selection Method for Object Detection13
Experimental results show that the proposed feature-selection method reduces dimensionality similar to PCA, but maintains the original measurement meanings while decreasing the computation time significantly.
- 13
- Event-based High-speed Ball Detection in Sports Video12
Experimental results confirm that the proposed deep learning-based method for detecting balls using event cameras can detect balls that are undetectable in RGB frames and outperform existing methods that utilize event-based ball detection.
- TC-OCR: TableCraft OCR for Efficient Detection & Recognition of Table Structure & Content12
This research proposes an end-to-end pipeline that integrates deep learning models, including DETR, Cascade TabNet, and PP OCR v2, to achieve comprehensive image-based table recognition, showcasing a remarkable improvement of approximately 25% in the OCR Accuracy compared to the previous Table Transformer approach.
- 12
- Video Indexing, Search, Detection, and Description with Focus on TRECVID12
This tutorial presents and discusses some of the most important and fundamental content-based video retrieval problems such as recognizing predefined visual concepts, searching in videos for complex ad-hoc user queries, searching by image/video examples in a video dataset to retrieve specific objects, persons, or locations, detecting events, and finally bridging the gap between vision and language.
- Efficient Tracking of News Topics Based on Chronological Semantic Structures in a Large-Scale News Video Archive12
A structuring method for a news video archive, together with an interface that visualizes the structure, so that users could track the development of news topics according to their interest, efficiently and consider this as a first step towards facilitating video authoring by users based on existing contents in a large news video retrieval system.
- Knowledge propagation in large image databases using neighborhood information12
A neighbor-based influence propagation approach KProp is proposed which builds a voting model and propagates the knowledge associated to some objects to similar objects and shows that KProp can perform efficiently through matrix computations and achieve better performance with fewer labeled examples per object.
- Integrating multi-modal content analysis and hyperbolic visualization for large-scale news video retrieval and exploration12
A novel scheme to achieve more effective analysis, retrieval and exploration of large-scale news video collections by performing multi-modal video content analysis and synchronization and a novel hyperbolic visualization scheme is incorporated to visualize large- scale news topics according to their associations and interestingness.
- Key Image Extraction from a News Video Archive for Visualizing Its Semantic Structure12
A method to extract the key images for visualizing the semantic structure of news video archive by analyzing semantic structure using both images and texts is introduced.
- 12
- Certified Zeroth-order Black-Box Defense with Robust UNet Denoiser11
This work proposes a robust UNet denoiser (RDUNet) that ensures the robustness of black-box models trained on high-dimensional datasets and proposes a novelblack-box denoised smoothing (DS) defense mechanism, ZO-RUDS, by prepending the authors' RDUNet to the black- box model, ensuring black-boxes defense.
- 11
- Vehicle Re-Identification with Spatio-Temporal Model Leveraging by Pose View Embedding11
A two-branch framework for vehicle Re-ID is designed, including a Keypoint-based Pose Embedding Visual (KPEV) model and a Key pointed Pose-Guided Spatio-Temporal (KPGST) model, which is integrated into the framework and the results of KPEV and KPGST are fused based on a Bayesian network.
- Effective and Efficient: Toward Open-world Instance Re-identification11
This tutorial would be useful for multimedia content analysis and system-level multimedia retrieval, especially for an effective and efficient open-world ReID system for the practical, large-scale, and open-set domain.
- 11
- An interaction model between human and system for intuitive graphical search interface11
A new interaction model is proposed by defining four user elements namely user knowledge, information needs, thinking, and feelings, and five system elements namely system knowledge, interaction algorithm, knowledge base, retrieval algorithm, and database which provides real interactions like human conversations.
- Bag of visual words model for videos segmentation into scenes11
This paper automatically segment the videos into scenes (visually related video shots) and calculates similarity computation efficient as the similarity of each shot is computed with its L neighbors only, instead of whole pool of shots.
- Q-CSLBP: Compression of CSLBP Descriptor11
A framework to reduce the descriptor length (dimensions) upto 50% without applying any dimensionality reduction techniques such as PCA or LDA is evaluated and the effectiveness of proposed compressed descriptor (Q-CSLBP) with CSLBP, SIFT and PCA-SIFT is compared.
- Human gesture recognition using 3.5-dimensional trajectory features for hands-free user interface11
A new human motion recognition technique for a hands-free user interface that uses a depth image in addition to a normal grayscale image from a time-of-flight camera that measures the depth to objects, so various motions are accurately recognized.
- Face Retrieval in Broadcasting News Video by Fusing Temporal and Intensity Information11
First, representative faces are quickly generated by using facial intensities to organize the face dataset into clusters, and temporal information is introduced to reorganize cluster memberships so as to improve overall retrieval performance.
- 11
- Examination and enhancement of a ring-structured graphical search interface based on usability testing11
The interactive retrieval functionality of the Concentric Ring View, a ring structure-based graphical user interface for image retrieval with multi-faceted metadata, was evaluated according to a series of usability studies and two functionalities were added, personal history for reuse and relevance feedback.
- Multi-Stage Approach to Fast Face Detection11
By combining AdaBoost and SVM classifiers, the final system can obtain both fast and robust detection because most of the non-face patterns are quickly rejected in earlier layers, while only a small number of promising face patterns are robustly classified in the later layers.
- Fusion of local and global features for efficient object detection11
A two-stage method for efficient object detection that combines full advantages of AdaBoost and SVM to achieve a reasonable balance in both training and classification speed and investigates the proposed method to detect different kinds of objects.
- 11
- 10
- Towards Ball Spin and Trajectory Analysis in Table Tennis Broadcast Videos via Physically Grounded Synthetic-to-Real Transfer10
To the best of the knowledge, this work is the first to present a method for spin and trajectory prediction in simple monocular broadcast videos, achieving an accuracy of 92.0% in spin classification and a 2D reprojection error of 0.19% of the image diagonal.
- A Culturally-diverse Multilingual Multimodal Video Benchmark & Model10
A multilingual Video LMM benchmark, named ViMUL-Bench, to evaluate Video LMMs across 14 languages, a machine translated multilingual video training set comprising 1.2 million samples and a simple multilingual video LMM, named ViMUL, are introduced to provide a better tradeoff between high-and low-resource languages for video understanding.
- 10
- Only a Few Classes Confusing: Pixel-Wise Candidate Labels Disambiguation for Foggy Scene Understanding10
This paper replaces the one-hot pseudo label with a candidate label set (CLS) that consists of only a few ambiguous classes and exploits its effects on self-training-based unsupervised domain adaptation to disambiguate the confusing semantics.
- 10
- Progressive Domain Adaptation for Robot Vision Person Re-identification10
This paper demonstrates a guiding robot with person followers system, which recognizes the follower using a person re-identification technology, and first adopts existing face recognition and person tracking methods to generate person tracklets with different IDs.
- Beyond Intra-modality Discrepancy: A Comprehensive Survey of Heterogeneous Person Re-identification.10
The state-of-the-art methods comprehensively with respect to four main application scenarios -- low-resolution, infrared, sketch and text are reviewed.
- 10
- [Paper] Image Retrieval with Fisher Vectors of Binary Features10
This paper derives the closed-form approximation of the Fisher vector of binary features which are modeled by the Bernoulli mixture model and shows that the Fisher vectors representation improves the accuracy of image retrieval by 25% compared with a bag of binary words approach.
- 10
- Multimedia Event Detection Using Segment-Based Approach for Motion Feature10
This paper proposes using a segment-based approach for video representation, which means original videos are divided into segments for feature extraction and classification, while still keeping the evaluation at the video level.
- NII-UIT-VBS: A Video Browsing Tool for Known Item Search10
A video browsing tool for the known item search task to reduce the number of segments by several ways such as applying visual filters and skimming representative keyframes and a coarse-to-fine based approach is employed to quickly find the target clip.
- Construction of image retrieval systems focused on user knowledge interaction10
The objective was to apply different kinds of database with the proposed graphical search interface, and to verify the effectiveness focused on user knowledge structure in searching because it allowed users to easily modify received information to suitable knowledge.
- 10
- High-Speed Detection of Intruders from Image Sequences Taken by Rotating Camera10
A high-speed detection method of intruders from image sequences that are taken by a rotating camera that uses the Viola-Jones detection that utilizes AdaBoost as one of the Boosting techniques and makes comparison between its method and existing two methods that are able to detect moving objects in real-time.
- DisCO: Portrait Distortion Correction with Perspective-Aware 3D GANs9
This work first performs 3D GAN inversion using a perspective-distorted input facial image by jointly optimizing the intrinsic and extrinsic camera parameters and the face latent code, and develops starting from a short distance, optimization scheduling, reparametrizations, and geometric regularization.
- Fundus Image Classification and Retinal Disease Localization with Limited Supervision9
A new method is proposed that shows evidential heatmaps by DCNN to ophthalmologists, and then obtained their feedback of selecting images with reasonable evidential regions, and improves accuracy from 90.1% to 94.5% in comparison with the existing method.
- Towards robots reasoning about group behavior of museum visitors: Leader detection and group tracking9
A computer vision algorithm is introduced which is able to detect and track a leader within a group of people – the science communicator, for this particular case – and distinguish between group members and non-group members as well, all by means of a cognitive and logical behaviour analysis of their interactions on scene.
- 9
- Deep Multi-label Hashing for Large-Scale Visual Search Based on Semantic Graph9
Extensive experiments over several large image datasets confirm that the proposed DMLH method outperforms state-of-the-art supervised and unsupervised image retrieval approaches, with a gain ranging from 6.25% to 38.9% in terms of mean average precision.
- Efficient large-scale multi-class image classification by learning balanced trees9
Experimental results on large-scale datasets including Caltech-256, SUN-397, ILSVRC2010-1K, and ImageNet-10K, show that the proposed label-tree based method outperforms other state-of-the-art methods.
- NHK STRL at TRECVID 2010: Semantic Indexing and Surveillance Event Detection.9
This paper uses random forests instead of the support vector machine (SVM) widely used in existing techniques as a classification algorithm to reduce the computational time needed for training anddetection.
- 9
- 9
- 8
- Density-Ratio Based Personalised Ranking from Implicit Feedback8
This study aims to establish an approach to learn personalised ranking from implicit feedback, which reconciles the training efficiency of the pointwise approach and ranking effectiveness of the pairwise counterpart.
- 8
- 8
- 8
- NII-UIT at MediaEval 2016 Predicting Media Interestingness Task.8
An overview of the framework used in MediaEval 2016 for the PMI task and the performance results for both subtasks of predicting image and video interestingness are discussed.
- 8
- 8
- 8
- 8
- 8
- 8
- 8
- A New Type of Video Scene Classification System Based on Typical Model Database8
A new type approach based on typical scene model dat,abase is proposed, which can classify the scene of image without recognition of object and the user does not have to prepare the knowledge for recognition of objects and scenes.
- 8
- 7
- 7
- From Scarcity to Understanding: Transfer Learning for the Extremely Low Resource Irish Sign Language7
- 7
- 7
- 7
- Large-Scale R-CNN with Classifier Adaptive Quantization7
This paper presents a novel quantization method designed forlinear classification wherein the quantization error is re-defined for linear classification and approximates the error as the empirical error with pre-defined multiple exemplar classifiers and captures the variance and common attributes of object category classifiers effectively.
- 7
- NII-UIT: A Tool for Known Item Search by Sequential Pattern Filtering7
This paper presents an interactive tool for searching a known item in a video or a video archive using query patterns formulated by users for filtering, which can deal with both visual and descriptive known item search.
- 7
- 7
- News Shot Cloud: Ranking TV News Shots by Cross TV-Channel Filtering for Efficient Browsing of Large-Scale News Video Archives7
A visualization method of TV news shots using the popularity-based filtering across multiple TV channels, which can be regarded as social filtering by TV broadcasters or popularity ranking among TV channels is presented.
- Automatic Pitch Type Recognition System from Single-View Video Sequences of Baseball Broadcast Videos7
A system that automatically recognizes individual pitch types like screwballs and sliders in baseball broadcast videos using only the same single-view broadcast baseball videos used by the human specialists, and accordingly a number of features are used based on the advice of those specialists.
- Commercial film detection and identification based on a dual-stage temporal recurrence hashing algorithm7
A dual-stage temporal recurrence hashing algorithm for fully unsupervised and super-fast Commercial Film (CF) mining in large-scale broadcast video archives is proposed and performed CF detection and identification on the 1-month sequence in merely 87 minutes.
- Large scale news video database browsing and retrieval via information visualization7
A novel framework is developed to enable more effective visual analysis and retrieval of large-scale news videos via interactive visualization, so that the audiences can find news stories of interest at first glance.
- 7
- 7
- Hierarchical Debiasing and Noisy Correction for Cross-domain Video Tube Retrieval6
The unsupervised domain adaptation model via Hierarchical dEbiAsing and noisy corRecTion (HEART) for cross-domain video tube retrieval is proposed, which contains two characteristic modules: Layered Feature Debiasing and Pseudo Label Refinement.
- Personalized Fashion Recommendation Using Pairwise Attention6
This paper introduces a personalized fashion recommendation system based on high-dimensional input of user-and environment information that can improve the performance up to 53.29% over the comparison method on MSE, mAP, and Recall.
- 6
- Video instance search via spatial fusion of visual words and object proposals6
This work proposes a novel fusion method to significantly boost the accuracy of instance search systems and proposes a new re-ranking scheme with three weighting functions corresponding to the three categories of visual word pairs to compute the final similarity score between a query topic and a video shot.
- 6
- NII-UIT Browser: A Multimodal Video Search System6
In this work, an interactive system for searching a known scene in a video database can be described and searched by its visual cues or audio genres, and a large number of possible queries can be generated to find the matchedscene in a short time.
- 6
- 6
- 6
- Efficient concept detection by fusing simple visual features6
This paper introduces a general framework for efficient and scalable concept detection by fusing SVM classifiers trained by only simple visual features such as color moments, edge orientation histogram and local binary patterns.
- New Approach for Hierarchical Classifier Training and Multi-level Image Annotation6
A novel hierarchical boosting algorithm is proposed by incorporating concept ontology and multi-task learning to achieve hierarchical image classifier training to enable automatic multi-level image annotation.
- 6
- Large-scale news video retrieval via visualization6
A novel framework is proposed to help audiences browse and retrieve news video clips without the need of keywords and interesting keyframes and keywords are automatically extracted from News video clips and visually represented according to their interestingness and informativeness measurement.
- 6
- 5
- 5
- RanLayNet: A Dataset for Document Layout Detection used for Domain Adaptation and Generalization5
A versatile dataset capable of training models with robustness and adaptability to diverse document formats is developed, and it is demonstrated that a deep layout identification model trained on this dataset exhibits enhanced performance compared to a model trained solely on actual documents.
- Towards Causality Inference for Very Important Person Localization5
A causal inference framework is established that unpacks the causes of previous methods and derives a new principled solution for VIPLoc, which treats the scene as confounding factor, allowing the ever-elusive confounding effects to be eliminated and the essential determinants to be uncovered.
- 5
- An Interactive Design for Visualizable Person Re-Identification5
It is demonstrated that a system offering these visual representations can effectively expedite and improve a person re-identification analysis and make it a much user-friendly experience.
- 5
- 5
- 5
- Person Re-identification Using Deformable Part Models5
A machine learning-based method to extract body parts that is based on Deformable Part Models (DPM), which is easy to train and has robust performance and experiments show that the proposed method outperforms state of the art methods.
- 5
- NII, Japan at MediaEval 2011 violent scenes detection task5
A comprehensive evaluation of performance of visual feature representations for MediaEval 2011 Violent Scenes Detection Task using keypoint detectors and SIFT and COLOR SIFT as descriptor of the region around these keypoints.
- 5
- 5
- Rushes summarization using different redundancy elimination approaches5
In the first approach, one keyframe is used for representing a shot when doing clustering; and sub-segments are selected using the motion information for generating the summary.
- 5
- 5
- Person X Detector.5
An automatic Person X detector using frontal faces in videos solely is designed and the architecture of person X detector and the evaluation results are illustrated.
- 5
- 5
- Towards actor/actress identification in drama videos5
After the system is trained using an episode of a drama series by given several tens of manual annotations to face sequences, it can automatically annotate face sequences not only of the trained episode, but of the other episodes of the drama series.
- Uplifting Table Tennis: A Robust, Real-World Application for 3D Trajectory and Spin Estimation4
This work proposes a novel two-stage pipeline that divides the problem into a front-end perception task and a back-end 2D-to-3D uplifting task, and specifically re-engineer the uplifting model to be robust to common real-world artifacts, such as missing detections and varying frame rates.
- 4
- 4
- Video Action Retrieval Using Action Recognition Model4
This paper proposes a method to retrieve similar videos by focusing on people and their actions in the videos, and applies this method to various datasets and TV videos to demonstrate its usefulness.
- 4
- 4
- Cloud platform for deep learning-based CAD via collaboration between Japanese medical societies and institutes of informatics4
A cloud platform for collecting multi-modal large- scale medical images from hospitals through medical societies and conducting various Deep Learning-based CAD research via collaboration between Japanesemedical societies and institutes of informatics is built.
- Rephrasing Visual Questions by Specifying the Entropy of the Answer Distribution4
A novel task, rephrasing the questions by controlling the ambiguity of the questions, and an interesting observation that it is harder to increase than to reduce ambiguity.
- Exploring Temporal Communities in Mass Media Archives4
This work proposes to extract the backbone of dynamic graphs in order to ease community detection and guide the exploration of trends evolution, and explores the role of soccer in 6 years of TV/radio magazines in France and North Korea in about 10 years of Japanese news.
- 4
- 4
- 4
- Learning Balanced Trees for Large Scale Image Classification4
The experimental results on the datasets such as Caltech-256, SUN-397, and ImageNet-1K show that the classification accuracy of the proposed approach outperforms that of other state of the art methods.
- User Emotion Sensing in Search Process based on Chromatic Sensation4
A model for sensing user emotion on the basis of the psychological principle, namely, color sensation is proposed, which extracts the present feeling of the user by using the colors browsed or selected by the user.
- Nagoya University at TRECVID 2014: the Instance Search Task.4
This paper takes the asymmetrical dissimilarities based system, which performed best in the INS2013 task, as the baseline, and re-rank with an improved spatial verification method, and shows that, the re-ranking algorithm is able to further improve the baseline system at a rather fast speed.
- Violent scene detection using mid-level feature4
The results show that the proposed method outperforms the standard approach directly using low-level features, and also narrows the semantic gap between low- level features and violence concept.
- 4
- Automated real-time video surveillance summarization framework4
A real-time video surveillance summarization framework intended for minimizing the time requirement for time critical tasks, based on compact moving objects in time–space, and dynamic region adaptation was used as the background subtraction algorithm to select the best foreground for each object.
- 4
- NII, Japan at ImageCLEF 2011 Photo Annotation Task.4
The participation of NII, Japan in the ImageCLEF 2011 Photo Annotation Task using NII-KAORI-SECODE framework is described and the performance of global features and local features for this task is evaluated.
- 4
- News Topic Tracking and Re-ranking with Query Expansion Based on Near-Duplicate Detection4
A novel scheme of mining topic-related stories through a query-expansion algorithm on the basis of near duplicates built on top of text that outperformed traditional methods that only use textual features is proposed.
- PageRank with Text Similarity and Video Near-Duplicate Constraints for News Story Re-ranking4
A novel application of PageRank is proposed, which is a pseudo-relevance feedback algorithm, and the constraints built on top of text to improve the relevance quality, and explores the pair-wise constraints derived from video near-duplicates for constraint-driven re-ranking.
- Large-scale news topic tracking and key-scene ranking with video near-duplicate constraints4
A novel news topic tracking and re-ranking system for analyzing both text similarity and video near-duplicate constraints and a proposed simple but effective query-expansion algorithm for improving the representativeness of a search query.
- Video rushes summarization utilizing retake characteristics4
A novel means of representative image selection for comparison which can be applied to remove redundancy is introduced which makes best use of normal behavior of shooting i.e., being motionless until start trigger, and after the trigger performer and camera move almost the same way (same motion.)
- Media objects for user-centered similarity matching4
Media objects for document description suited to images and videos are defined by integrating a user-centered definition of importance for similarity matching, which allows for meaningful similarity matching between document pairs and between users’ queries and documents.
- 4
- Concept Detection Using Local Binary Patterns and SVM.4
This work introduces a method for detecting concepts by using only visual information, in which the feature vector of each frame is formed by concatenating histograms of local binary patterns extracted from overlapped and sized rectangles within the frame.
- CLIPS-LSR-NII Experiments at TRECVID 2005.4
An original and promising framework for incorporating contextual information (from image content) into the concept detection process is pro-posed, which combines local and global clas-sifiers with stacking, using SVM.
- Generalized Histogram: Empirical Optimization of Low Dimensional Features for Image Matching4
The generalized Histogram is proposed as low-dimensional representation of an image for efficient and precise image matching, and achieves more than 90% precision and recall with 16-D generalized histogram compared to the ground truth computed by normalized cross correlation.
- 4
- 4
- 4
- 3
- An Entropy Clustering Approach for Assessing Visual Question Difficulty3
This work proposes a novel approach to identify clusters of difficult visual questions for Visual Question Answering (VQA) without direct supervision or annotations to the difficulty, and has the advantage of being able to assess the difficulty of visual questions without ground-truth by assigning them to one of the clusters.
- A multimedia document browser based on multilayer networks3
A visual analytics system that is designed to support search tasks in multimedia document archives and provides complex querying, semantic overviews of time, and visual, and textual concepts combined with analysis.
- Salient Time Slice Pruning and Boosting for Person-Scene Instance Search in TV Series3
This paper extends the focus from time point (single video shot) to time slice (multiple consecutive video shots) in the time-line, and evaluates the importance of salient time slices to boost the aggregation results.
- Evaluating Face Tracking for Political Analysis in Japanese News Over a Long Period of Time3
This paper compares the two face tracking methods as well as a traditional text-based method by using the same dataset of Japanese broadcasting news spanning over a decade and demonstrates that the modern implementation outperforms the two other methods.
- Group Re-identification via Transferred Single and Couple Representation Learning.3
A Transferred Single and Couple Representation Learning Network is proposed, which treats a group as multiple persons and transfer the domain of a labeled ReID dataset to a G-ReID target dataset style to learn single representations and takes into account the neighborhood relationship in a group.
- FaceCloud3
Inspired by word clouds, these clouds allow to grasp visual and textual semantic information captured from a multimedia collection all at once, and then enables direct access to the relevant video.
- 3
- 3
- MANet3
This paper proposes a novel Modal Attention Network (MANet), which extends the standard encoder-decoder network by adapting the attention mechanism to video modalities and emphasizes the impact of each modality with respect to the word to be generated.
- P-S Instance Retrieval via Early Elimination and Late Expansion3
An early elimination and late expansion method is proposed to improve the accuracy of P-S Instance Retrieval, and experimental results on the large-scale TRECVID INS dataset demonstrate the effectiveness of the proposed method.
- Query-Adaptive R-CNN for Open-Vocabulary Object Detection and Retrieval.3
This work proposes Query-Adaptive R-CNN, a simple yet strong framework for open-vocabulary object detection that outperforms all state-of-the-art methods in the Flickr30k Entities phrase localization task and proposes negative phrase augmentation, a generic approach for exploiting hard negatives in the training of open- vocabulary object detection.
- Semantic Extraction and Object Proposal for Video Search3
A sketch based search system which detects and indexes many objects proposed by an object proposal algorithm which not only leverage the concepts but also the spatial relations between them.
- 3
- 3
- 3
- Scalable approaches for content based video retrieval3
The goal is to develop approaches which require lowest annotation cost or computational cost while achieving competitive accuracy so that they can facilitate building scalable and comprehensive video retrieval systems.
- An Application Search Interface Including Sense-related Search Facets3
A ring-structured graphical search interface for multi-faceted metadata to solve two reported usability problems by preparing dials for each facet and dividing the functions for ring rotation into two operations.
- The Future of Multimedia Analysis and Mining: Visions from the Shonan Meeting3
The meeting was really interesting, and the participants had a fun time with an Kamakura excursion and fine dinners, in addition to in-depth discussions on ready-to-go hot research topics.
- 3
- 3
- NTT Communication Science Laboratories and NII in TRECVID 2010 Instance Search Task3
Several existing image retrieval methods in combination are employed in combination (local feature matching, region matching, and global feature matching), according to the type and property of query images to locate query topics from a collection of reference videos.
- A Novel Retake Detection Using LCS and SIFT Algorithm3
The results support the notion that the proposed method provides a reasonably high degree of accuracy, and identifies the likely causes of poor accuracy for further improvements.
- NHK STRL at TRECVID 2009: Surveillance Event Detection and High-Level Feature Extraction.3
NHK Science and Technology Research Laboratories participated in two tasks at TRECVID 2009: surveillance event detection task and high level feature extraction task and adopt the BVM method to reduce the computational cost of training.
- Proposal of an Information Compilation Method for Massive News Video Data Based on their Time-Series Semantic Structure3
A browsing and editing interface that enables the user to browse through news stories along the topic thread structure, and also assists the compilation of selected news stories as a customized video summary or a documentary is introduced.
- Video search by multi-modal and clustering analysis3
A video search system using multi-modal and clustering analysis that retrieves results by keywords or by visual contents such as frames or rectangle regions within a frame and only representative frames of these clusters are shown to users.
- ReSeDis: A Dataset for Referring-based Object Search across Large-Scale Image Collections2
This work introduces Referring Search and Discovery (ReSeDis), the first task that unifies corpus-level retrieval with pixel-level grounding, and curates a benchmark in which every description maps uniquely to object instances scattered across a large, diverse corpus, eliminating unintended matches.
- Foundation Model Insights and a Multi-Model Approach for Superior Fine-Grained One-shot Subset Selection2
Surprising insights are uncovered: FMs consistently outperform traditional IEs on fine-grained datasets, whereas their advantage diminishes on coarse-grained datasets with noisy labels.
- LoA-Trans: Enhancing Visual Grounding by Location-Aware Transformers2
A query selection mechanism that generates possible object locations other than the estimated center and use them as location-aware queries as a remedy for possible inaccurate center estimation, and introduces a TaskSyn Network in the de-coder to better coordination between REC and RES.
- CNNGen: A Generator and a Dataset for Energy-Aware Neural Architecture Search2
This work contributes to energy-aware NAS with a grammar-based Convolutional Neural Network generator producing diverse architectures not based on a reference one, and three state-of-the-art predictors releasing the need for trained models for performance and energy estimation.
- The Effects of Short Video-Sharing Services on Video Copy Detection2
The effects of short video-sharing services on VCD are examined by constructing a dataset that has short video-sharing service characteristics and it is seen that segment-level VCD in short video-sharing services is more difficult than those in general video-sharing services and video alignment component mainly suppress the detection performance in short video-sharing services.
- 2
- Improving Generalization of Metric Learning via Listwise Self-distillation2
This work proposes a simple yet effective regularization, namely Listwise Self-Distillation (LSD), which progressively distills a model's own knowledge to adaptively assign a more appropriate distance target to each sample pair in a batch.
- Efficient Nearest Neighbor Search by Removing Anti-hub2
This work empirically found that such unnecessary vectors have low hubness scores and thus can be easily identified beforehand and removed by removing anti-hubs, achieving a memory-efficient search while preserving accuracy.
- Analysis of Typefaces Designed for Readers with Developmental Dyslexia2
Developmental dyslexia is a specific learning disability that is characterized by severe difficulties in learning to read, and amongst various supporting technologies, there are typefaces specially designed for readers with Dyslexia.
- 2
- Efficient Large Scale Image Classification via Prediction Score Decomposition2
This work analyzes the correlation among classes, suppresses redundancy, and generates a small number of classifiers that best approximate the prediction scores of the original large number of classes to achieve a good trade-off between testing efficiency and classification accuracy.
- Video Event Detection by Exploiting Word Dependencies from Image Captions2
This paper proposes a new approach to obtain the relationship between concepts by exploiting the syntactic dependencies between words in the image captions, which significantly reduces the number of informative combinations between concepts.
- Query Bootstrapping: A Visual Mining Based Query Expansion2
An adaptive support algorithm that adaptively determines both the minimum support and maximum support by referring to the first round’s retrieval list is proposed that gives significant performance boosts over that of state-of-the-art QE on three benchmark datasets.
- 2
- 2
- Bidirectional extraction and recognition of scene text with layout consistency2
This paper proposes a method for recognizing text by utilizing the layout consistency of a text string, and calls this two-way process—from extraction and recognition to layout, and from layout to extraction and Recognition—“bidirectional” to discriminate it from previous feedback refinement approaches.
- 2
- Tell Me about TV Commercials of This Product2
An interactive system on top of video retrieval in a large collection of three-year five-channel TV commercial videos, which is to the best of the knowledge, the largest commercial data set used for retrieval so far.
- 2
- [Paper] Using Trajectory Features to Recognize Human Actions Within Crowd Sequences of Real Surveillance Video2
This paper proposes action recognition methods based on multiple trajectories that can identify human actions within crowd sequences of real surveillance video and discusses how to select the proper method to detect actions in crowd situations through an analysis of these experimental results.
- 2
- A video navigation interface using multi-faceted search hierarchies2
A graphical search interface for multi-faceted hierarchical metadata and a video navigation system implementing this 'Revolving Cube Show' interface, enabling users to search flexibly and intuitively by using simple operations to combine attributes.
- 2
- 2
- Improving Retake Detection by Adding Motion Feature2
This work proposes another approach to improve the performance of retake detection by combining the motion feature extracted from each sub-sequence and the features extracting from each representing keyframe, which is effective on the standard benchmark dataset of TRECVID BBC Rushes 2007.
- 2
- Using Visual-Textual Mutual Information and Entropy for Inter-modal Document Indexing2
The proposed information theoretic model is implemented, and the results of experiments assessing its performance on two collections show that information theory is an interesting framework to automatically annotate documents.
- 2
- Mining Large-Scale News Video Database Via Knowledge Visualization2
A novel framework to enable intuitive mining and exploration of large-scale video news databases via knowledge visualization and a statistical video analysis algorithm is proposed to extract the semantics from the video reports.
- Semantic Retrieval in a Large-Scale Video Database by Using Both Image and Text Feature2
A new retrieval method of intergrating both image features and text informations derived from video data is demonstrated, which not only image similiarity, but also narrow the retrieval sets in advance by employing searching in text keywords.
- 2
- Toward The MEdiaSys VIdeo Search Engine (MEVISE)2
The innovative architecture of MEVISE, the MediaSys VIdeo Search Engine over a large scale network is described and the content-based representation of the video data inside the MediaSY server is introduced.
- 1
- AD2AT: Audio Description to Alternative Text, a Dataset of Alternative Text from Movies1
This paper introduces AD2AT, a dataset of alt text derived from professionally tailored audio descriptions in movies, and demonstrates the limitations of state-of-the-art image captioning and text generation models in producing effective alt text.
- 1
- Computer-Assisted Noise Pareidolia Tests Through Patient Emulation1
This research presents a machine learning-based framework to assist in assessing the pareidolia phenomenon, and shows a performance superior to baseline methods in distinguishing pareidolia from other similar visual illusions.
- Fairness without Labels: Pseudo-Balancing for Bias Mitigation in Face Gender Classification1
P pseudo-balancing consistently improves fairness while preserving or enhancing accuracy and suggests that even in the absence of label supervision, access to a demographically balanced or moderately skewed unlabeled dataset can serve as a powerful resource for debiasing existing computer vision models.
- 1
- 1
- Pose-aware Outfit Transfer between Unpaired in-the-wild Fashion Images1
This paper develops a method that generates realistic try-on images with unpaired images from in-the-wild datasets by combining a reconstruction and a content loss to preserve the detail and style of the transferred clothing, background and the target person.
- 1
- 1
- 1
- 1
- 1
- NTT Communication Science Laboratories and National Institute of Informatics at TRECVID 2017 Instance Search.1
The approaches that were tested in the TRECVID 2017 Instance Search (INS) task, where shots including a person at a location were to be retrieved after a location query and person query were given, used ASMK for the location search and OpenFace for the person search.
- Information Retrieval Model using Generalized Pareto Distribution and Its Application to Instance Search1
The proposed information retrieval model corresponds to the extension of the divergence from independence and is designed to be data-driven and the effectiveness is experimentally confirmed.
- Bayesian Exponential Inverse Document Frequency and Region-of-Interest Effect for Enhancing Instance Search Accuracy1
This paper first analyzes the discriminative power in the Best Match (BM) 25 formula and provides its calculation method from the Bayesian point of view and justifies the EIDF by calculating the discrimInative Power in the BM25 from theBayesian viewpoint.
- 1
- NII-HITACHI-UIT at TRECVID 2015 instance search1
Two methods to improve last year instance search framework are proposed, based on post processing scheme that try to rerank top K shots returned from BOW model and a query-adaptive weighting technique between DPM object detectors score and BOW's score.
- Insight DCU at TRECVID 20151
Insight-DCU participated in the instance search (INS), semantic indexing (SIN), and localization tasks (LOC) this year, and it was found that using features from the convolutional layers improved performance over Features from the fully connected layers used in previous years, and that weak geometric consistency improves performance for local feature ranking.
- Efficient Cross-Domain Image Retrieval by Multi-Level Matching and Spatial Verification for Structural Similarity1
A multi-level matching process which dynamically selects the most suitable matching constraint for feature matching and adopt a fast spatial verification to describe the structure similarity for cross-domain image retrieval.
- A Classification-Based Approach for Retake and Scene Detection in Rushes Video1
Instead of quantizing each video segment into a label, this work formulate it as a binary classification problem between pairs of segments, and uses this information as the input for the Smith-Waterman algorithm to detect and group similar video sub-sequences to find retakes.
- Connect commercial films with realities1
Many known state-of-the-art retrieval technologies are integrated into a commercial film retrieval system, which manages to index a huge commercial dataset archived from five TV channels within recent three years.
- An interactive exploration system that visually supports learning of country features1
A country data exploration system as a teaching tool for geography education at the elementary- and middle-school levels that enables students to search for information flexibly and intuitively by combining attributes using simple operations.
- Improving Image Categorization by Using Multiple Instance Learning with Spatial Relation1
Discriminative Multiple Instance Learning is used for image categorization when a label is provided for the entire training image only instead of the object region and a subset of sub-windows per positive bag is selected.
- 1
- 1
- The python computer vision framework1
PyCVF allows rapid development of applications and it provides standardized tools for common operations such as browsing datasets, applying transformations to one dataset on-the-fly, computing features, indexing multimedia datasets, querying for nearest-neighbors, training a statistical model, or browsing the result in a 3d-space.
- 1
- Image-based quiz generation from news video archives based on principal object1
The attempt to generate quizzes automatically from news programs, as novel method for content creation by formulating the engineering problem of quiz generation - selection of images suitable for a quiz, selection of text that explains the image, and the generation of alternative choices that are similar but different by selecting from subtitles.
- 1
- 1
- 1
- Advances in multimedia information processing - PCM 2004 : 5th Pacific Rim Conference on Multimedia, Tokyo, Japan, November 30 - December 3, 2004 : proceedings1
Human-Scale Virtual Reality and Interaction, WARAJI: Foot-Driven Navigation Interfaces for Virtual Reality Applications, and Action Generation from Natural Language.
- 1
- 1
- A Syntactical Approach to the Database Construction Method from Document Images.1
This paper presents a database construCtion method fromJOurnalcontentimagesfocuslngOntheloglCalstruCtureconstruction using the knowledge-directedapproach of AdatabaseschemarepresentsaloglCalstructure of data.
- Understanding Rule Generation Supporting System for Drawing Understanding using Interaction with User.1
It is nrcessary 7 0 artlieve rfftcirnt and powrrT11I drawing irnderstandinpl in orrlcr to r~alize l a r ~ e SCRIP ptactiral drawing image datalmes.
- Data model generation in image database systems1
The important aspects of a next-generation image database with advanced capabilities suitable for use with a variety of practical applications and also for modeling mixed image data are discussed, and a new method for model generation is described, emphasizing automatic model acquisition.
- –
- –
- DiverXplorer: Stock Image Exploration via Diversity Adjustment for Graphic Design–
An image exploration prototype that enables stepwise adjustment of diversity, allowing users to transition from diverse overviews to increasingly focused subsets during exploration, and aims to provide a novel perspective on how to implement transitions between diversity and similarity.
- –
- –
- –
- How to Teach Video Retrieval Models: A Systematic Study of Multi-Teacher Knowledge Distillation–
This work examines teacher application strategies and shows that per-batch teacher scheduling, which assigns a single teacher to each mini-batch during training, provides a more stable distillation paradigm, and provides practical design guidelines for constructing stable and optimization-aware multi-teacher distillation pipelines for text-to-video retrieval.
- –
- –
- –
- Fairness Auditing of Limited-Access Models Using Protected Datasets–
This framework enables practical, privacy-preserving fairness auditing for real-world scenarios involving regulatory and certification contexts where mutual distrust between auditors and service providers creates operational constraints.
- –
- –
- Evaluating VQA Models’ Consistency in the Scientific Domain–
Two approaches are proposed that slightly modify the image-question pair without changing the question’s meaning to gain a deeper comprehension of VQA models’ question understanding: choices permutation and question rephrasing.
- –
- MIS '24: 1st ACM Multimedia Workshop on Multi-modal Misinformation Governance in the Era of Foundation Models–
This workshop brings together researchers and practitioners to discuss key topics in multi-modal misinformation governance, including new datasets, evaluation techniques, ethical considerations, methodological progress, case studies, and future research directions.
- Mitigating robust overfitting via self-residual-calibration regularization (Abstract Reprint)–
The proposed Self-Residual-Calibration (SRC) calculates the absolute residual between adversarial and natural logit features corresponding to the ground-truth labels and utilizes the pinball loss to minimize the quantile residual between them, resulting in more robust regularization.
- –
- –
- –
- Self-distillation with Online Diffusion on Batch Manifolds Improves Deep Metric Learning–
This paper proposes a simple but effective Progressive Self-Distillation (PSD), which distills the knowledge progressively from the model itself during training, and extends it with an Online Batch Diffusion Process (OBDP), which is to capture the local geometric structure of manifolds in each batch, so that it can reveal the intrinsic relationships among samples in the batch and produce better soft distance targets.
- Scalable Personalised Item Ranking through Parametric Density Estimation–
A learning-to-rank approach, which achieves convergence speed comparable to the pointwise counterpart while performing similarly to the pairwise counterpart in terms of ranking effectiveness and efficiency, and is equivalent to an IRGAN variant under a certain condition.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Introduction to Special Issue of the 23rd International Conference on Multimedia Modeling (MMM 2017)–
This special issue of Multimedia Tools and Applications contains selected papers from the 23rd International Conference on Multimedia Modeling (MMM 2017), which was held in Reykjavík, Iceland, on January 4–6, 2017.
- Speech Reconstitution using Multi-view Silent Videos.–
This work encompasses the boundaries of multimedia research by putting forth a model which leverages silent video feeds from multiple cameras recording the same subject to generate intelligent speech for a speaker.
- –
- –
- Tracked Instance Search–
Experimental results show that the instance search system improves substantially its performance when using tracking, and a study on the improvement of the system when using multiple independent instances of the same person is presented.
- Face2Graph: Base de données graphe et visualisation pour l'annotation d'archives vidéos–
L’exploration du réseau de co-apparition avec Face2Graph propose une manière originale d’annoter les visages dans notre archive.
- –
- –
- –
- Introduction of New Associate Editors–
I would like to welcome 22 new outstanding researchers who are joining the IEEE Transactions on Circuits and Systems for Video Technology (TCSVT) Editorial Board (EB) as Associate Editors.
- Human Action Recognition from Depth Videos Using Pool of Multiple Projections with Greedy Selection–
The proposed greedy method outperforms other state-of-the-art methods while using a small number of projections and best complementary projections will be considered first when searching for optimal combination of projections.
- Self-localization and Navigation in Dynamic Search Hierarchy for Video Retrieval Interface–
The improved interface improved on the previously proposed graphical search interface, ‘Revolving Cube Show,’ for multi-faceted metadata, enabling users to search flexibly and intuitively by using simple operations to combine attributes.
- –
- –
- –
- News Archive Exploration Combining Face Detection and Tracking with Network Visual Analytics–
It is proposed to combine the two domains of visual analytics and data analytics by demonstrating a visual exploration of over a decade of the news program from the Japanese broadcaster NHK News 7, and derive social networks from face detection and tracking of this large dataset.
- –
- –
- Cross-View Action Recognition by Projection-Based Augmentation–
This paper takes the advantages of depth data to augment viewpoints from an initial camera viewpoint to tackle the problem based on exploiting the discrimination in action execution through various viewpoints and shows the important role of viewpoints to improve the performance of action recognition.
- –
- –
- –
- –
- –
- Enhanced Visualization of News Shot Cloud with Employing Circular Layout–
This poster presents an enhancement method of the news shot cloud by modifying the thumbnail layout from the conventional linear layout to the circular, which is more helpful to identify informative news shots rapidly, and at the same time, it also helps users to examine less-weighted news shots to find lower-ranked but unexpectedly informative ones.
- PVSS–
A Portable Visual Search Service for a researcher who desires to make his or her algorithm appear to be as "cool" as the best interactive demos at image retrieval conferences, which handles the user interaction with the audience and gives researchers more freedom to explain their algorithm.
- –
- –
- –
- Recommend-Me–
This paper presents a novel recommendation system, named Recommend-Me, which automatically recommends good query regions to users and proposes and applies a novel branch-and-bound based algorithm to efficiently identify region pairs with highest scores.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- NHK STRL at TRECVID 2011: Surveillance Event Detection and Semantic Indexing–
A novel local feature method that considers the spatial relationship between feature points and the co-occurrence frequency of feature descriptors at each feature point to reduce the computational time required for training and detection.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- CLIPS and NII at TRECvid: Shot segmentation and feature extraction–
An original and promising framework for in- corporating contextual information (from image content) into the concept detection process, which combines local and global clas- sifiers with stacking, using SVM.
- –
- –
- –
- Face Sequence Matching with Certainty Factor Evaluation–
This paper describes a method of robust face sequence matching with rejection mechanism, which is expected that faces with not enough information for matching are excepted by the rejection mechanism taking advantage of certainty factor evaluation.
- an actor/actress annotation system for drama videos–
An semi-automated actor/actress annotation system for drama videos, using a newly developed method for robust face sequence extraction and matching, and face classification is described.
- –
- –
- –
- –
- –
- Robust Line Drawing Understanding Incorporating Efficient Closed Symbols Extraction.–
A graph search technique is used for the symbols extraction method to provide efficient model prediction and verification facilities, and quantitative consideration on the computational cost is shown.
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.