Multi-annotator learning (MAL) aims to model annotator-specific labeling patterns. However, existing methods face a critical challenge: they simply skip updating annotator-specific model parameters when encountering missing labels—a common scenario in real-world crowdsourced datasets where each annotator labels only small subsets of samples. This leads to inefficient data utilization and overfitting risks. To this end, we propose a novel similarity-weighted semi-supervised learning framework (SimLabel) that leverages inter-annotator similarities to generate weighted soft labels for missing annotations, enabling the utilization of unannotated samples rather than skipping them entirely. We further introduce a confidence-based iterative refinement mechanism that combines maximum probability with entropy-based uncertainty to prioritize predicted high-quality pseudo-labels to impute missing labels, jointly enhancing similarity estimation and model performance over time. For evaluation, we contribute a new multimodal multi-annotator dataset, AMER2, with high and more variable missing rates, reflecting real-world annotation sparsity and enabling evaluation across different sparsity levels. Extensive experiments validate the effectiveness of our method.

Multi-annotator learning traditionally aggregates diverse annotations to approximate a single “ground truth”, treating disagreements as noise. However, this paradigm faces fundamental challenges: subjective tasks often lack absolute ground truth, and sparse annotation coverage makes aggregation statistically unreliable. We introduce a paradigm shift from sample-wise aggregation to annotator-wise behavior modeling. By treating annotator disagreements as valuable information rather than noise, modeling annotator-specific behavior patterns can reconstruct unlabeled data to reduce annotation cost, enhance aggregation reliability, and explain annotator decision behavior. To this end, we propose QuMAB (Query-based Multi-Annotator Behavior Pattern Learning), which uses lightweight queries to model individual annotators while capturing inter-annotator correlations as implicit regularization, preventing overfitting to sparse individual data while maintaining individualization and improving generalization, with a visualization of annotator focus regions offering an explainable analysis of behavior understanding. We contribute two large-scale datasets with dense per-annotator labels: STREET (4,300 labels/annotator) and AMER (average 3,118 labels/annotator), the first multimodal multi-annotator dataset. Extensive experiments demonstrate the superiority of our QuMAB in modeling individual annotators’ behavior patterns, their utility for consensus prediction, and applicability under sparse annotations.

Large Language Models (LLMs) have demonstrated significant capabilities, particularly in the domain of question answering (QA). However, their effectiveness in QA is often undermined by the vagueness of user questions. To address this issue, we introduce single-round instance-level prompt optimization, referred to as question rewriter. By enhancing the intelligibility of human questions for black-box LLMs, our question rewriter improves the quality of generated answers. The rewriter is optimized using direct preference optimization based on feedback collected from automatic criteria for evaluating generated answers; therefore, its training does not require costly human annotations. The experiments across multiple black-box LLMs and long-form question answering (LFQA) datasets demonstrate the efficacy of our method. This paper provides a practical framework for training question rewriters and sets a precedent for future explorations in prompt optimization within LFQA tasks.

Recent approaches for visually-rich document understanding (VrDU) uses manually annotated semantic groups, where a semantic group encompasses all semantically relevant but not obviously grouped words. As OCR tools are unable to automatically identify such grouping, we argue that current VrDU approaches are unrealistic. We thus introduce a new variant of the VrDU task, real-world visually-rich document understanding (ReVrDU), that does not allow for using manually annotated semantic groups. We also propose a new method, ReLayout, compliant with the ReVrDU scenario, which learns to capture semantic grouping through arranging words and bringing the representations of words that belong to the potential same semantic group closer together. Our experimental results demonstrate the performance of existing methods is deteriorated with the ReVrDU task, while ReLayout shows superiour performance.

Gender bias in vision-language foundation models (VLMs) raises concerns about their safe deployment and is typically evaluated using benchmarks with gender annotations on real-world images. However, as these benchmarks often contain spurious correlations between gender and non-gender features, such as objects and backgrounds, we identify a critical oversight in gender bias evaluation: Do spurious features distort gender bias evaluation? To address this question, we systematically perturb non-gender features across four widely used benchmarks (COCO-gender, FACET, MIAP, and PHASE) and various VLMs to quantify their impact on bias evaluation. Our findings reveal that even minimal perturbations, such as masking just 10% of objects or weakly blurring backgrounds, can dramatically alter bias scores, shifting metrics by up to 175% in generative VLMs and 43% in CLIP variants. This suggests that current bias evaluations often reflect model responses to spurious features rather than gender bias, undermining their reliability. Since creating spurious feature-free benchmarks is fundamentally challenging, we recommend reporting bias metrics alongside feature-sensitivity measurements to enable a more reliable bias assessment.

The real value of knowledge lies not just in its accumulation, but in its potential to be harnessed effectively to conquer the unknown. Although recent multimodal large language models (MLLMs) exhibit impressing multimodal capabilities, they often fail in rarely encountered domain-specific tasks due to limited relevant knowledge. To explore this, we adopt visual game cognition as a testbed and select Monster Hunter: World as the target to construct a multimodal knowledge graph (MH-MMKG), which incorporates multi-modalities and intricate entity relations. We also design a series of challenging queries based on MH-MMKG to evaluate the models' ability for complex knowledge retrieval and reasoning. Furthermore, we propose a multi-agent retriever that enables a model to autonomously search relevant knowledge without additional training. Experimental results show that our approach significantly enhances the performance of MLLMs, providing a new perspective on multimodal knowledge-augmented reasoning and laying a solid foundation for future research.

Large-scale vision-language models, such as CLIP, are known to contain societal bias regarding protected attributes (e.g., gender, age). This paper aims to address the problems of societal bias in CLIP. Although previous studies have proposed to debias societal bias through adversarial learning or test-time projecting, our comprehensive study of these works identifies two critical limitations: 1) loss of attribute information when it is explicitly disclosed in the input and 2) use of the attribute annotations during debiasing process. To mitigate societal bias in CLIP and overcome these limitations simultaneously, we introduce a simple-yet-effective debiasing method called SANER (societal attribute neutralizer) that eliminates attribute information from CLIP text features only of attribute-neutral descriptions. Experimental results show that SANER, which does not require attribute annotations and preserves original information for attribute-specific descriptions, demonstrates superior debiasing ability than the existing methods.

In this paper, we introduce a novel taskfor video understanding that focuses on detecting editing intentions in political advertisement videos. Political advertisement videos are edited with some intentions (e.g., “associating some candidates with negative emotions”) of making people unthinkingly believe the messages in the videos, potentially ending up with some irrational bias. Detecting such intentions is thus the primary step toward fairer decision-making based on the messages themselves. To this end, we classify such editing intentions into 10 categories (referred to as communication techniques) in consultation with a professional editor as well as based on communication techniques presented in the natural language processing community, and build a dataset of 12,526 political advertisement videos, each of which is annotated with several communication technique segments. We also explore the capability of existing video understanding models in detecting editing intentions over the dataset, which identifies new dimensions of challenges to be addressed.

Object detection in art is a valuable tool for the digital humanities, as it allows for faster identification of objects in artistic and historical images compared to humans. However, annotating such images poses significant challenges due to the need for specialized domain expertise. We present NADA (no annotations for detection in art), a pipeline that leverages diffusion models' art-related knowledge for object detection in paintings without the need for full bounding box supervision. Our method, which supports both weakly-supervised and zero-shot scenarios and does not require any fine-tuning of its pretrained components, consists of a class proposer based on large vision-language models and a class-conditioned detector based on Stable Diffusion. NADA is evaluated on two artwork datasets, ArtDL 2.0 and IconArt, outperforming prior work in weakly-supervised detection, while being the first work for zero-shot object detection in art. Code is available at https://github.com/patrick-john-ramos/nada

Abstract Analysis of cultural heritages, particularly their construction years, provides new insights into human history. However, due to natural disasters, wars, material deterioration, and human errors, records documenting the construction years of many artifacts have often been lost. Historians and experts can estimate construction years within specific ranges using chemical-based analysis technologies or extensive historical research. Given the vast number of collected artifacts, applying these conventional methods to every artifact is impractical. To address this challenge, we developed a deep neural network model designed for Buddha statues to estimate an artifact’s construction year from its image. One major challenge in this task is the heterogeneity of the labels: the training samples include both precise construction years and possible ranges (e.g., a dynasty or a century) estimated by historians. To unify these heterogeneous labels during training, we represent them as probabilistic distributions. In our previous work Qian et al. (2021), we assumed that the ambiguity in heterogeneous construction year labels followed a Gaussian distribution, assigning the highest likelihood to the midpoint of the designated time range. However, this assumption does not always hold. In this paper, we propose representing heterogeneous construction year labels as a uniform distribution, assigning equal probability to all points within the designated time range. Based on this label representation, we designed a semi-supervised learning loss function to leverage both labeled and unlabeled samples during training. Our experimental results demonstrate that our method achieves a mean absolute error of 34.3 years on a test set consisting of Buddha statues constructed between 400 and 1403. These results are further analyzed in two ways. First, we compared our model’s performance to the image quality BRISQUE score, revealing a correlation between higher image quality and lower prediction error rates. Second, we validated our predictions with experts, assessing the level of agreement with our model, the challenges in determining construction years, and identifying features of interest in the artifacts.

Abstract The use of AI systems for managing large-scale cultural heritage artifacts has become possible due to the rise of digitization. To classify such content, machine learning is typically used, where contextual information is important to structure the data. One way to capture context is through a knowledge graph. In this study, we propose a newd graph neural networks, we can improve artwork classification by utilizing the relationships between entities of the knowledge graph. Our experiments demonstrate that this approach achieves state-of-the-art results on multiple classification tasks across three datasets (SemArt paintings, Buddha statues, and Ukiyo-e woodblock prints). Moreover, our approach is effective in dealing with unbalanced data and we explore the use of both graph attention mechanisms and focal loss functions.

Social images are often associated with rich but noisy tags from community contributions. Although social tags can potentially provide valuable semantic training information for image retrieval, existing studies all fail to effectively filter noises by exploiting the cross-modal correlation between image content and tags. The current cross-modal vision-and-language representation learning methods, which selectively attend to the relevant parts of the image and text, show a promising direction. However, they are not suitable for social image retrieval since: (1) they deal with natural text sequences where the relationships between words can be easily captured by language models for cross-modal relevance estimation, while the tags are isolated and noisy; (2) they take (image, text) pair as input, and consequently cannot be employed directly for unimodal social image retrieval. This paper tackles the challenge of utilizing cross-modal interactions to learn precise representations for unimodal retrieval. The proposed framework, dubbed CGVR (Cross-modal Guided Visual Representation), extracts accurate semantic representations of images from noisy tags and transfers this ability to image-only hashing subnetwork by a carefully designed training scheme. To well capture correlated semantics and filter noises, it embeds a priori common-sense relationship among tags into attention computation for joint awareness of textual and visual context. Experiments show that CGVR achieves approximately 8.82 and 5.45 points improvement in MAP over the state-of-the-art on two widely used social image benchmarks. CGVR can serve as a new baseline for the image retrieval community. The code is provided at https://github.com/zhaowanqing/CGVR.

Several studies have raised awareness about social biases in image generative models, demonstrating their predisposition towards stereotypes and imbalances. This paper contributes to this growing body of research by introducing an evaluation protocol that analyzes the impact of gender indicators at every step of the generation process on Stable Diffusion images. Leveraging insights from prior work, we explore how gender indicators not only affect gender presentation but also the representation of objects and layouts within the generated images. Our findings include the existence of differences in the depiction of objects, such as instruments tailored for specific genders, and shifts in overall layouts. We also reveal that neutral prompts tend to produce images more aligned with masculine prompts than their feminine counterparts. We further explore where bias originates through representational disparities and how it manifests in the images via prompt-image dependencies, and provide recommendations for developers and users to mitigate potential bias in image generation.

We investigate the impact of deep generative models on potential social biases in upcoming computer vision mod-els. As the internet witnesses an increasing influx of AI-generated images, concerns arise regarding inherent biases that may accompany them, potentially leading to the dis-semination of harmful content. This paper explores whether a detrimental feedback loop, resulting in bias amplification, would occur if generated images were used as the training data for future models. We conduct simulations by progressively substituting original images in COCO and CC3M datasets with images generated through Stable Dif-fusion. The modified datasets are used to train OpenCLIP and image captioning models, which we evaluate in terms of quality and bias. Contrary to expectations, our findings indicate that introducing generated images during training does not uniformly amplify bias. Instead, instances of bias mitigation across specific tasks are observed. We further explore the factors that may influence these phenomena, such as artifacts in image generation (e.g., blurry faces) or pre-existing biases in the original datasets.

Yusuke Hirota, Jerone Andrews, Dora Zhao, Orestis Papakyriakopoulos, Apostolos Modas, Yuta Nakashima, Alice Xiang. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Large language models (LLMs) have enhanced the capacity of vision-language models to caption visual text.This generative approach to image caption enrichment further makes textual captions more descriptive, improving alignment with the visual context.However, while many studies focus on the benefits of generative caption enrichment (GCE), are there any negative side effects?We compare standardformat captions and recent GCE processes from the perspectives of "gender bias" and "hallucination", showing that enriched captions suffer from increased gender bias and hallucination.Furthermore, models trained on these enriched captions amplify gender bias by an average of 30.9% and increase hallucination by 59.5%.This study serves as a caution against the trend of making captions more descriptive.Code is available at https://github.com/rebnej/ GenCapEnrich-bias.

We introduce an emotional stimuli retrieval task that targets extracting emotional regions that evoke people's emotions (i.e., emotional stimuli) in artworks. This task offers new challenges to the community because of the diversity of artwork styles and the subjectivity of emotions, which can be a suitable testbed for benchmarking the capability of the current neural networks to deal with human emotion. For this task, we construct a dataset called APOLO for quantifying emotional stimuli retrieval performance in artworks by crowd-sourcing pixel-level annotation of emotional stimuli. APOLO contains 6,781 emotional stimuli in 4,718 artworks for validation and testing. We also evaluate eight baseline methods, including a dedicated one, to show the difficulties of the task and the limitations of the current techniques through qualitative and quantitative experiments. Our data and methods are available in https://github.com/Tianwei3989/apolo.

In this companion paper, we provide the artifacts of the GOYA model for disentangling content and style in art paintings, as presented at ICMR2023. The scripts are written in Python.

Large language models (LLMs) have recently showcased remarkable capabilities, spanning a wide range of tasks and applications, including those in the medical domain. Models like GPT-4 excel in medical question answering but may face challenges in the lack of interpretability when handling complex tasks in real clinical settings. We thus introduce the diagnostic reasoning dataset for clinical notes (DiReCT), aiming at evaluating the reasoning ability and interpretability of LLMs compared to human doctors. It contains 511 clinical notes, each meticulously annotated by physicians, detailing the diagnostic reasoning process from observations in a clinical note to the final diagnosis. Additionally, a diagnostic knowledge graph is provided to offer essential knowledge for reasoning, which may not be covered in the training data of existing LLMs. Evaluations of leading LLMs on DiReCT bring out a significant gap between their reasoning ability and that of human doctors, highlighting the critical need for models that can reason effectively in real-world clinical scenarios.

Contrastive image representation learning through instance discrimination has shown impressive transfer performance. Recent strategies have focused on pushing the limit of their transfer performance for dense prediction tasks, particularly when conducting pre-training on scene images with complex structures. Initial approaches employ pixel-level contrastive pre-training to optimize dense spatial features, while subsequent methods utilize region-mining algorithms to capture holistic regional semantics and address the issue of semantically inconsistent scene image crops. In this paper, we revisit pixel-level contrastive pre-training on scene images. Contrary to the assumption that pixel-level learning falls short in achieving these objectives, we demonstrate its under-explored potentials: (1) it can effectively learn holistic regional semantics more simply compared to region-level methods, and (2) it intrinsically provides tools to mitigate the impact of semantically inconsistent views involved with scene-level training images. We propose PixCon, a pixel-level contrastive learning framework, and explore two variants with different positive matching strategies to investigate the potential of pixel-level learning. Additionally, when PixCon incorporates a novel semantic reweighting approach tailored for scene image pre-training, it outperforms or matches the performance of previous region-level methods in object detection and semantic segmentation tasks across multiple benchmarks.1

Large-scale models trained on extensive datasets, have emerged as the preferred approach due to their high generalizability across various tasks. In-context learning (ICL), a popular strategy in natural language processing, uses such models for different tasks by providing instructive prompts but without updating model parameters. This idea is now being explored in computer vision, where an input-output image pair (called an in-context pair) is supplied to the model with a query image as a prompt to exemplify the desired output. The efficacy of visual ICL often depends on the quality of the prompts. We thus introduce a method coined Instruct Me More (InMeMo), which augments in-context pairs with a learnable perturbation (prompt), to explore its potential. Our experiments on mainstream tasks reveal that InMeMo surpasses the current state-of-the-art performance. Specifically, compared to the baseline without learnable prompt, InMeMo boosts mIoU scores by 7.35 and 15.13 for foreground segmentation and single object detection tasks, respectively. Our findings suggest that InMeMo offers a versatile and efficient way to enhance the performance of visual ICL with lightweight training. Code is available at https://github.com/Jackieam/InMeMo.

The increasing tendency to collect large and uncurated datasets to train vision-and-language models has raised concerns about fair representations. It is known that even small but manually annotated datasets, such as MSCOCO, are affected by societal bias. This problem, far from being solved, may be getting worse with data crawled from the Internet without much control. In addition, the lack of tools to analyze societal bias in big collections of images makes addressing the problem extremely challenging. Our first contribution is to annotate part of the Google Conceptual Captions dataset, widely used for training vision-and-language models, with four demographic and two contextual attributes. Our second contribution is to conduct a comprehensive analysis of the annotations, focusing on how different demographic groups are represented. Our last contribution lies in evaluating three prevailing vision-and-language tasks: image captioning, text-image CLIP embeddings, and text-to-image generation, showing that societal bias is a persistent problem in all of them. https://github.com/noagarcia/phase

Image captioning models are known to perpetuate and amplify harmful societal bias in the training set. In this work, we aim to mitigate such gender bias in image captioning models. While prior work has addressed this problem by forcing models to focus on people to reduce gender mis-classification, it conversely generates gender-stereotypical words at the expense of predicting the correct gender. From this observation, we hypothesize that there are two types of gender bias affecting image captioning models: 1) bias that exploits context to predict gender, and 2) bias in the probability of generating certain (often stereotypical) words because of gender. To mitigate both types of gender biases, we propose a framework, called LIBRA, that learns from synthetically biased samples to decrease both types of biases, correcting gender misclassification and changing gender-stereotypical words to more neutral ones.

Human evaluation is critical for validating the performance of text-to-image generative models, as this highly cognitive process requires deep comprehension of text and images. However, our survey of 37 recent papers reveals that many works rely solely on automatic measures (e.g., FID) or perform poorly described human evaluations that are not reliable or repeatable. This paper proposes a standardized and well-defined human evaluation protocol to facilitate verifiable and reproducible human evaluation in future works. In our pilot data collection, we experimentally show that the current automatic measures are incompatible with human perception in evaluating the performance of the text-to-image generation results. Furthermore, we provide insights for designing human evaluation experiments reliably and conclusively. Finally, we make several resources publicly available to the community to facilitate easy and fast implementations.

Interpreting and explaining the behavior of deep neural networks is critical for many tasks. Explainable AI provides a way to address this challenge, mostly by providing per-pixel relevance to the decision. Yet, interpreting such explanations may require expert knowledge. Some recent attempts toward interpretability adopt a concept-based framework, giving a higher-level relationship between some concepts and model decisions. This paper proposes Bottleneck Concept Learner (BotCL), which represents an image solely by the presence/absence of concepts learned through training over the target task without explicit supervision over the concepts. It uses self-supervision and tailored regularizers so that learned concepts can be human-understandable. Using some image classification tasks as our testbed, we demonstrate BotCL's potential to rebuild neural networks for better interpretability11Code is avaliable at https://github.com/wbw520/BotCL and a simple demo is available at https://botcl.liangzhili.com/.

Recently, there has been an increased interest in cross-data research problems, such as predicting air quality using life logging images, predicting congestion using weather and tweets data, and predicting sleep quality using daily exercises and meals. Although several research focusing on multimodal data analytics have been performed, few studies have been conducted on cross-data research (e.g., cross-modal data, cross-domain, cross-platform). The article collection “Intelligent Cross-Data Analysis and Retrieval” aims to encourage research in intelligent cross-data analytics and retrieval and contribute to the creation of a sustainable society. Researchers from diverse domains such as well-being, disaster prevention and mitigation, mobility, climate, tourism and healthcare are welcome to contribute to this Research Topic.

The duality of content and style is inherent to the nature of art. For humans, these two elements are clearly different: content refers to the objects and concepts in the piece of art, and style to the way it is expressed. This duality poses an important challenge for computer vision. The visual appearance of objects and concepts is modulated by the style that may reflect the author’s emotions, social trends, artistic movement, etc., and their deep comprehension undoubtfully requires to handle both. A promising step towards a general paradigm for art analysis is to disentangle content and style, whereas relying on human annotations to cull a single aspect of artworks has limitations in learning semantic concepts and the visual appearance of paintings. We thus present GOYA, a method that distills the artistic knowledge captured in a recent generative model to disentangle content and style. Experiments show that synthetically generated images sufficiently serve as a proxy of the real distribution of artworks, allowing GOYA to separately represent the two elements of art while keeping more information than existing methods.

Fake news detection in social media has become increasingly important due to the rapid proliferation of personal media channels and the consequential dissemination of misleading information. Existing methods, which primarily rely on multimodal features and graph-based techniques, have shown promising performance in detecting fake news. However, they still face a limitation, i.e., sparsity in graph connections, which hinders capturing possible interactions among tweets. This challenge has motivated us to explore a novel method that densifies the graph's connectivity to capture denser interaction better. Our method constructs a cross-modal tweet graph using CLIP, which encodes images and text into a unified space, allowing us to extract potential connections based on similarities in text and images. We then design a Feature Contextualization Network with Label Propagation (FCN-LP) to model the interaction among tweets as well as positive or negative correlations between predicted labels of connected tweets. The propagated labels from the graph are weighted and aggregated for the final detection. To enhance the model's generalization ability to unseen events, we introduce a domain generalization loss that ensures consistent features between tweets on seen and unseen events. We use three publicly available fake news datasets, Twitter, PHEME, and Weibo, for evaluation. Our method consistently improves the performance over the state-of-the-art methods on all benchmark datasets and effectively demonstrates its aptitude for generalizing fake news detection in social media.

Video summarization aims to select the most informative subset of frames in a video to facilitate efficient video browsing. Unsupervised methods usually rely on heuristic training objectives such as diversity and representativeness. However, such methods need to bootstrap the online-generated summaries to compute the objectives for importance score regression. We consider such a pipeline inefficient and seek to directly quantify the frame-level importance with the help of contrastive losses in the representation learning literature. Leveraging the contrastive losses, we propose three metrics featuring a desirable key frame: local dissimilarity, global consistency, and uniqueness. With features pre-trained on the image classification task, the metrics can already yield high-quality importance scores, demonstrating competitive or better performance than past heavily-trained methods. We show that by refining the pre-trained features with a lightweight contrastively learned projection module, the frame-level importance scores can be further improved, and the model can also leverage a large number of random videos and generalize to test videos with decent performance.

Abstract Few-shot learning (FSL) approaches, mostly neural network-based, assume that pre-trained knowledge can be obtained from base (seen) classes and transferred to novel (unseen) classes. However, the black-box nature of neural networks makes it difficult to understand what is actually transferred, which may hamper FSL application in some risk-sensitive areas. In this paper, we reveal a new way to perform FSL for image classification, using a visual representation from the backbone model and patterns generated by a self-attention based explainable module. The representation weighted by patterns only includes a minimum number of distinguishable features and the visualized patterns can serve as an informative hint on the transferred knowledge. On three mainstream datasets, experimental results prove that the proposed method can enable satisfying explainability and achieve high classification results. Code is available at https://github.com/wbw520/MTUNet .

Abstract Humor can be induced by various signals in the visual, linguistic, and vocal modalities emitted by humans. Finding humor in videos is an interesting but challenging task for an intelligent system. Previous methods predict humor in the sentence level given some text (e.g., speech transcript), sometimes together with other modalities, such as videos and speech. Such methods ignore humor caused by the visual modality in their design, since their prediction is made for a sentence. In this work, we first give new annotations to humor based on a sitcom by setting up temporal segments of ground truth humor derived from the laughter track. Then, we propose a method to find these temporal segments of humor. We adopt an approach based on sliding window, where the visual modality is described by pose and facial features along with the linguistic modality given as subtitles in each sliding window. We use long short-term memory networks to encode the temporal dependency in poses and facial features and pre-trained BERT to handle subtitles. Experimental results show that our method improves the performance of humor prediction.

We study societal bias amplification in image captioning. Image captioning models have been shown to perpetuate gender and racial biases, however, metrics to measure, quantify, and evaluate the societal bias in captions are not yet standardized. We provide a comprehensive study on the strengths and limitations of each metric, and propose LIC, a metric to study captioning bias amplification. We argue that, for image captioning, it is not enough to focus on the correct prediction of the protected attribute, and the whole context should be taken into account. We conduct extensive evaluation on traditional and state-of-the-art image captioning models, and surprisingly find that, by only focusing on the protected attribute prediction, bias mitigation models are unexpectedly amplifying bias.

Mean Average Precision (mAP) is the primary evaluation measure for object detection. Although object detection has a broad range of applications, mAP evaluates detectors in terms of the performance of ranked instance retrieval. Such the assumption for the evaluation task does not suit some downstream tasks. To alleviate the gap between downstream tasks and the evaluation scenario, we propose Optimal Correction Cost (OC-cost), which assesses detection accuracy at image level. OC-cost computes the cost of correcting detections to ground truths as a measure of accuracy. The cost is obtained by solving an optimal transportation problem between the detections and the ground truths. Unlike mAp, OC-cost is designed to penalize false positive and false negative detections properly, and every image in a dataset is treated equally. Our experimental result validates that OCscost has better agreement with human preference than a ranking-based measure, i.e., mAP for a single image. We also show that detectors' rankings by OC-cost are more consistent on different data splits than mAP. Our goal is not to replace mAP with OC-cost but provide an additional tool to evaluate detectors from another aspect. To help future researchers and developers choose a target measure, we provide a series of experiments to clarify how mAP and OC-cost differ.

Evaluation measures have a crucial impact on the direction of research. Therefore, it is of utmost importance to develop appropriate and reliable evaluation measures for new applications where conventional measures are not well suited. Video Moment Retrieval (VMR) is one such application, and the current practice is to use R@K,$\theta$for evaluating VMR systems. However, this measure has two disadvantages. First, it is rank-insensitive: It ignores the rank positions of successfully localised moments in the top-K ranked list by treating the list as a set. Second, it binarizes the Intersection over Union (IoU) of each retrieved video moment using the threshold$\theta$and thereby ignoring fine-grained localisation quality of ranked moments. We propose an alternative measure for evaluating VMR, called Average Max IoU (AxIoU), which is free from the above two problems. We show that AxIoU satisfies two important axioms for VMR evaluation, namely, Invariance against Redundant Moments and Monotonicity with respect to the Best Moment, and also that R@ K,$\theta$satisfies the first axiom only. We also empirically examine how Ax-IoU agrees with R@K,$\theta$, as well as its stability with respect to change in the test data and human-annotated temporal boundaries.

Body language such as conversational gesture is a powerful way to ease communication. Conversational gestures do not only make a speech more lively but also contain semantic meaning that helps to stress important information in the discussion. In the field of robotics, giving conversational agents (humanoid robots or virtual avatars) the ability to properly use gestures is critical, yet remain a task of extraordinary difficulty. This is because given only a text as input, there are many possibilities and ambiguities to generate an appropriate gesture. Different to previous works we propose a new method that explicitly takes into account the gesture types to reduce these ambiguities and generate human-like conversational gestures. Key to our proposed system is a new gesture database built on the TED dataset that allows us to map a word to one of three types of gestures: “Imagistic” gestures, which express the content of the speech, “Beat” gestures, which emphasize words, and “No gestures.” We propose a system that first maps the words in the input text to their corresponding gesture type, generate type-specific gestures and combine the generated gestures into one final smooth gesture. In our comparative experiments, the effectiveness of the proposed method was confirmed in user studies for both avatar and humanoid robot.

We annotate 35,000 SNS posts with both the writer’s subjective sentiment polarity labels and the reader’s objective ones to construct a Japanese sentiment analysis dataset. Our dataset includes intensity labels (none, weak, medium, and strong) for each of the eight basic emotions by Plutchik (joy, sadness, anticipation, surprise, anger, fear, disgust, and trust) as well as sentiment polarity labels (strong positive, positive, neutral, negative, and strong negative). Previous studies on emotion analysis have studied the analysis of basic emotions and sentiment polarity independently. In other words, there are few corpora that are annotated with both basic emotions and sentiment polarity. Our dataset is the first large-scale corpus to annotate both of these emotion labels, and from both the writer’s and reader’s perspectives. In this paper, we analyze the relationship between basic emotion intensity and sentiment polarity on our dataset and report the results of benchmarking sentiment polarity classification.

We have witnessed the rise of cross-data against multimodal data problems recently. The cross-modal retrieval system uses a textual query to look for images; the air quality index can be predicted using lifelogging images; the congestion can be predicted using weather and tweets data; daily exercises and meals can help to predict the sleeping quality are some examples of this research direction. Although vast investigations focusing on multimodal data analytics have been developed, few cross-data (e.g., cross-modal data, cross-domain, cross-platform) research has been carried on. In order to promote intelligent cross-data analytics and retrieval research and to bring a smart, sustainable society to human beings, the specific article collection on "Intelligent Cross-Data Analysis and Retrieval" is introduced. This Research Topic welcomes those who come from diverse research domains and disciplines such as well-being, disaster prevention and mitigation, mobility, climate change, tourism, healthcare, and food computing

Visually grounded paraphrases (VGPs) are different phrasal expressions describing the same visual concept in an image. Previous studies treat VGP identification as a binary classification task, which ignores various phenomena behind VGPs (i.e., different linguistic interpretation of the same visual concept) such as linguistic paraphrases and VGPs from different aspects. In this paper, we propose semantic typology for VGPs, aiming to elucidate the VGP phenomena and deepen the understanding about how human beings interpret vision with language. We construct a large VGP dataset that annotates the class to which each VGP pair belongs according to our typology. In addition, we present a classification model that fuses language and visual features for VGP classification on our dataset. Experiments indicate that joint language and vision representation learning is important for VGP classification. We further demonstrate that our VGP typology can boost the performance of visually grounded textual entailment.

Conventional 3D convolutional neural networks (CNNs) are computationally expensive, memory intensive, prone to overfitting, and most importantly, there is a need to improve their feature learning capabilities. To address these issues, we propose spatio-temporal short term Fourier transform (STFT) blocks, a new class of convolutional blocks that can serve as an alternative to the 3D convolutional layer and its variants in 3D CNNs. An STFT block consists of non-trainable convolution layers that capture spatially and/or temporally local Fourier information using a STFT kernel at multiple low frequency points, followed by a set of trainable linear weights for learning channel correlations. The STFT blocks significantly reduce the space-time complexity in 3D CNNs. In general, they use 3.5 to 4.5 times less parameters and 1.5 to 1.8 times less computational costs when compared to the state-of-the-art methods. Furthermore, their feature learning capabilities are significantly better than the conventional 3D convolutional layer and its variants. Our extensive evaluation on seven action recognition datasets, including Something-something v1 and v2, Jester, Diving-48, Kinetics-400, UCF 101, and HMDB 51, demonstrate that STFT blocks based 3D CNNs achieve on par or even better performance compared to the state-of-the-art methods.

Have you ever looked at a painting and wondered what is the story behind it? This work presents a framework to bring art closer to people by generating comprehensive descriptions of fine-art paintings. Generating informative descriptions for artworks, however, is extremely challenging, as it requires to 1) describe multiple aspects of the image such as its style, content, or composition, and 2) provide background and contextual knowledge about the artist, their influences, or the historical period. To address these challenges, we introduce a multi-topic and knowledgeable art description framework, which modules the generated sentences according to three artistic topics and, additionally, enhances each description with external knowledge. The framework is validated through an exhaustive analysis, both quantitative and qualitative, as well as a comparative human evaluation, demonstrating outstanding results in terms of both topic diversity and information veracity.

Explainable artificial intelligence has been gaining attention in the past few years. However, most existing methods are based on gradients or intermediate features, which are not directly involved in the decision-making process of the classifier. In this paper, we propose a slot attention-based classifier called SCOUTER for transparent yet accurate classification. Two major differences from other attention-based methods include: (a) SCOUTER’s explanation is involved in the final confidence for each category, offering more intuitive interpretation, and (b) all the categories have their corresponding positive or negative explanation, which tells "why the image is of a certain category" or "why the image is not of a certain category." We design a new loss tailored for SCOUTER that controls the model’s behavior to switch between positive and negative explanations, as well as the size of explanatory regions. Experimental results show that SCOUTER can give better visual explanations in terms of various metrics while keeping good accuracy on small and medium-sized datasets. Code is available1.

We propose a new 2D pose refinement network that learns to predict the human bias in the estimated 2D pose. There are biases in 2D pose estimations that are due to differences between annotations of 2D joint locations based on annotators’ perception and those defined by motion capture (MoCap) systems. These biases are crafted into publicly available 2D pose datasets and cannot be removed with existing error reduction approaches. Our proposed pose refinement network allows us to efficiently remove the human bias in the estimated 2D poses and achieve highly accurate multi-view 3D human pose estimation.

In the e-learning context, how much the learner is concentrated and engaged, or the learners’ efficiency, is essential for providing adaptive and flexible materials, timely suggestions, etc., which can lead to efficient learning. In this work, we explore to predict learners’ efficiency with a realistic configuration, in which we use a webcam or a laptop PC’s built-in camera. Specifically, we first provide a feasible definition of the learners’ efficiency, and based on this definition, we predict one’s efficiency from facial behavior. We predict the learners’ efficiency using various convolutional neural networks. Results are discussed using different evaluation metrics.

The rise of digitization of cultural documents offers large-scale contents, opening the road for development of AI systems in order to preserve, search, and deliver cultural heritage. To organize such cultural content also means to classify them, a task that is very familiar to modern computer science. Contextual information is often the key to structure such real world data, and we propose to use it in form of a knowledge graph. Such a knowledge graph, combined with content analysis, enhances the notion of proximity between artworks so it improves the performances in classification tasks. In this paper, we propose a novel use of a knowledge graph, that is constructed on annotated data and pseudo-labeled data. With label propagation, we boost artwork classification by training a model using a graph convolutional network, relying on the relationships between entities of the knowledge graph. Following a transductive learning framework, our experiments show that relying on a knowledge graph modeling the relations between labeled data and unlabeled data allows to achieve state-of-the-art results on multiple classification tasks on a dataset of paintings, and on a dataset of Buddha statues. Additionally, we show state-of-the-art results for the difficult case of dealing with unbalanced data, with the limitation of disregarding classes with extremely low degrees in the knowledge graph.

Deep hashing has been widely used to approximate nearest-neighbor search for image retrieval tasks. Most of them are trained with image-label pairs without any inter-label relationship, which may not make full use of the real-world data. This paper presents deep hashing, named HA2SH, that leverages multiple types of labels with hierarchical structures that an ethnological museum assigns to their artifacts. We experimentally prove that HA2SH can learn to generate hashes that give a better retrieval performance. Our code is available at https://github.com/wbw520/minpaku.

Tomoyuki Kajiwara, Chenhui Chu, Noriko Takemura, Yuta Nakashima, Hajime Nagahara. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Humor is a very important communication tool; yet, it is an open problem for machines to understand humor. In this paper, we build a new multimodal dataset for humor prediction that includes subtitles and video frames, as well as humor labels associated with video's timestamps. On top of it, we present a model to predict whether a subtitle causes laughter. Our model uses the visual modality through facial expression and character name recognition, together with the verbal modality, to explore how the visual modality helps. In addition, we use an attention mechanism to adjust the weight for each modality to facilitate humor prediction. Interestingly, our experimental results show that the performance boost by combinations of different modalities, and the attention mechanism and the model mostly relies on the verbal modality.

We propose a novel video understanding task by fusing knowledge-based and video question answering. First, we introduce KnowIT VQA, a video dataset with 24,282 human-generated question-answer pairs about a popular sitcom. The dataset combines visual, textual and temporal coherence reasoning together with knowledge-based questions, which need of the experience obtained from the viewing of the series to be answered. Second, we propose a video understanding model by combining the visual and textual video content with specific knowledge about the show. Our main findings are: (i) the incorporation of knowledge produces outstanding improvements for VQA in video, and (ii) the performance on KnowIT VQA still lags well behind human accuracy, indicating its usefulness for studying current video modelling limitations.

In this paper, we propose a full pipeline of analysis of a large corpus about a century of public meeting in historical Australian news papers, from construction to visual exploration. The corpus construction method is based on image processing and OCR. We digitize and transcribe texts of the specific topic of public meeting. Experiments show that our proposed method achieves a F-score of 87.8% for corpus construction. As a result, we built a content search tool for temporal and semantic content analysis.

Retinal vessel segmentation is of great interest for diagnosis of retinal vascular diseases. To further improve the performance of vessel segmentation, we propose IterNet, a new model based on UNet [1], with the ability to find obscured details of the vessel from the segmented vessel image itself, rather than the raw input image. IterNet consists of multiple iterations of a mini-UNet, which can be 4× deeper than the common UNet. IterNet also adopts the weight-sharing and skip-connection features to facilitate training; therefore, even with such a large architecture, IterNet can still learn from merely 10~20 labeled images, without pre-training or any prior knowledge. IterNet achieves AUCs of 0.9816, 0.9851, and 0.9881 on three mainstream datasets, namely DRIVE, CHASE-DB1, and STARE, respectively, which currently are the best scores in the literature. The source code is available1.

Visual question answering (VQA) aims at answering questions about the visual content of an image or a video. Currently, most work on VQA is focused on image-based question answering, and less attention has been paid into answering questions about videos. However, VQA in video presents some unique challenges that are worth studying: it not only requires to model a sequence of visual features over time, but often it also needs to reason about associated subtitles. In this work, we propose to use BERT, a sequential modelling technique based on Transformers, to encode the complex semantics from video clips. Our proposed model jointly captures the visual and language information of a video scene by encoding not only the subtitles but also a sequence of visual concepts with a pretrained language-based Transformer. In our experiments, we exhaustively study the performance of our model by taking different input arrangements, showing outstanding improvements when compared against previous work on two well-known video VQA datasets: TVQA and Pororo.

Video summarization is a technique to create a short skim of the original video while preserving the main stories/content. There exists a substantial interest in automatizing this process due to the rapid growth of the available material. The recent progress has been facilitated by public benchmark datasets, which enable easy and fair comparison of methods. Currently the established evaluation protocol is to compare the generated summary with respect to a set of reference summaries provided by the dataset. In this paper, we will provide in-depth assessment of this pipeline using two popular benchmark datasets. Surprisingly, we observe that randomly generated summaries achieve comparable or better performance to the state-of-the-art. In some cases, the random summaries outperform even the human generated summaries in leave-one-out experiments. Moreover, it turns out that the video segmentation, which is often considered as a fixed pre-processing method, has the most significant impact on the performance measure. Based on our observations, we propose alternative approaches for assessing the importance scores as well as an intuitive visualization of correlation between the estimated scoring and human annotations.

Deep convolutional neural networks (CNNs) have established their feet in the ground of computer vision and machine learning, used in various applications. In this work, an attempt is made to learn a CNN for a task of facial expression recognition (FER). Our network has convolutional layers linked with an FC layer with a skip-connection to the classification layer. Motivation behind this design is that lower layers of a CNN are responsible for lower level features, and facial expressions can be mainly encoded in low-to-mid level features. Hence, in order to leverage the responses from lower layers, all convo-lutional layers are integrated via FC layers. Moreover, a network with shared parameters is used to extract landmark motion trajectory features. These visual and landmark features are fused to improve the performance. Our method is evaluated on the CK+ and Oulu-CASIA facial expression datasets.

Automatic art analysis aims to classify and retrieve artistic representations from a collection of images by using computer vision and machine learning techniques. In this work, we propose to enhance visual representations from neural networks with contextual artistic information. Whereas visual representations are able to capture information about the content and the style of an artwork, our proposed context-aware embeddings additionally encode relationships between different artistic attributes, such as author, school, or historical period. We design two different approaches for using context in automatic art analysis. In the first one, contextual data is obtained through a multi-task learning model, in which several attributes are trained together to find visual relationships between elements. In the second approach, context is obtained through an art-specific knowledge graph, which encodes relationships between artistic attributes. An exhaustive evaluation of both of our models in several art analysis problems, such as author identification, type classification, or cross-modal retrieval, show that performance is improved by up to 7.3% in art classification and 37.24% in retrieval when context-aware embeddings are used.

We introduce BUDA.ART, a system designed to assist researchers in Art History, to explore and analyze an archive of pictures of Buddha statues. The system combines different CBIR and classical retrieval techniques to assemble 2D pictures, 3D statue scans and meta-data, that is focused on the Buddha facial characteristics. We build the system from an archive of 50,000 Buddhism pictures, identify unique Buddha statues, extract contextual information, and provide specific facial embedding to first index the archive. The system allows for mobile, on-site search, and to explore similarities of statues in the archive. In addition, we provide search visualization and 3D analysis of the statues.

A paraphrase is a restatement of the meaning of a text in other words. Paraphrases have been studied to enhance the performance of many natural language processing tasks. In this paper, we propose a novel task iParaphrasing to extract visually grounded paraphrases (VGPs), which are different phrasal expressions describing the same visual concept in an image. These extracted VGPs have the potential to improve language and image multimodal tasks such as visual question answering and image captioning. How to model the similarity between VGPs is the key of iParaphrasing. We apply various existing methods as well as propose a novel neural network-based method with image attention, and report the results of the first attempt toward iParaphrasing.

Reconstruction of the shape and motion of humans from RGB-D is a challenging problem, receiving much attention in recent years. Recent approaches for full-body reconstruction use a statistic shape model, which is built upon accurate full-body scans of people in skin-tight clothes, to complete invisible parts due to occlusion. Such a statistic model may still be fit to an RGB-D measurement with loose clothes but cannot describe its deformations, such as clothing wrinkles. Observed surfaces may be reconstructed precisely from actual measurements, while we have no cues for unobserved surfaces. For full-body reconstruction with loose clothes, we propose to use lower dimensional embeddings of texture and deformation referred to as eigen-texturing and eigen-deformation, to reproduce views of even unobserved surfaces. Provided a full-body reconstruction from a sequence of partial measurements as 3D meshes, the texture and deformation of each triangle are then embedded using eigen-decomposition. Combined with neural-network-based coefficient regression, our method synthesizes the texture and deformation from arbitrary viewpoints. We evaluate our method using simulated data and visually demonstrate how our method works on real data.

Automatically generating a summary of a sports video poses the challenge of detecting interesting moments, or highlights, of a game. Traditional sports video summarization methods leverage editing conventions of broadcast sports video that facilitate the extraction of high-level semantics. However, user-generated videos are not edited and, thus, traditional methods are not suitable to generate a summary. In order to solve this problem, this paper proposes a novel video summarization method that uses players' actions as a cue to determine the highlights of the original video. A deep neural-network-based approach is used to extract two types of action-related features and to classify video segments into interesting or uninteresting parts. The proposed method can be applied to any sports in which games consist of a succession of actions. Especially, this paper considers the case of Kendo (Japanese fencing) as an example of a sport to evaluate the proposed method. The method is trained using Kendo videos with ground truth labels that indicate the video highlights. The labels are provided by annotators possessing a different experience with respect to Kendo to demonstrate how the proposed method adapts to different needs. The performance of the proposed method is compared with several combinations of different features, and the results show that it outperforms previous summarization methods.

The proliferation of off-the-shelf head-mounted displays (HMDs) let end-users enjoy virtual reality applications, some of which render a real-world scene using a novel view synthesis (NVS) technique. View-dependent texture mapping (VDTM) has been studied for NVS due to its photo-realistic quality. The VDTM technique renders a novel view by adaptively selecting textures from the most appropriate images. However, this process is computationally expensive because VDTM scans every captured image. For stereoscopic HMDs, the situation is much worse because we need to render novel views once for each eye, almost doubling the cost. This paper proposes light-weight VDTM tailored for an HMD. In order to reduce the computational cost in VDTM, our method leverages the overlapping fields of view between a stereoscopic pair of HMD images and pruning the images to be scanned. We show that the proposed method drastically accelerates the VDTM process without spoiling the image quality through a user study.

Augmented reality (AR) marker hiding is a technique to visually remove AR markers in a real-time video stream. A conventional approach transforms a background image with a homography matrix calculated on the basis of a camera pose and overlays the transformed image on an AR marker region in a real-time frame, assuming that the AR marker is on a planar surface. However, this approach may cause discontinuities in textures around the boundary between the marker and its surrounding area when the planar surface assumption is not satisfied. This paper proposes a method for AR marker hiding without discontinuities around texture boundaries even under nonplanar background geometry without measuring it. For doing this, our method estimates the dense motion in the marker's background by analyzing the motion of sparse feature points around it, together with a smooth motion assumption, and deforms the background image according to it. Our experiments demonstrate the effectiveness of the proposed method in various environments with different background geometries and textures.

Automatic sports video summarization poses the challenge of acquiring semantics of the original video, and existing work leverages various knowledge in application domains, e.g., structure of games and editing conventions. In this paper, we propose a personal sports video summarization method for self-recorded RGB-D videos, which became available to the public due to the commodification of off-the-shelf RGB-D sensors. We focus on sports whose games consist of a succession of actions and, unlike previous research, we use human action recognition on the depth sequences in order to acquire higher level semantics of the video. The recognition results are used along with an entropy-based activity measure to train a hidden Markov model of the highlights of different games to extract a summary from the original RGB-D video. We trained our novel highlights model with the subjective opinion of users with different experience in the sport. We took Kendo, a martial art, as an example sport to evaluate our method, and objectively/subjectively investigated the accuracy and quality of the generated summaries.

An enormous number of images are currently shared through social networking services such as Facebook. These images usually contain appearance of people and may violate the people's privacy if they are published without permission from each person. To remedy this privacy concern, visual privacy protection, such as blurring, is applied to facial regions of people without permission. However, in addition to image quality degradation, this may spoil the context of the image: If some people are filtered while the others are not, missing facial expression makes comprehension of the image difficult. This paper proposes an image melding-based method that modifies facial regions in a visually unintrusive way with preserving facial expression. Our experimental results demonstrated that the proposed method can retain facial expression while protecting privacy.

Recent popularization of camera devices, including action cams and smartphones, enables us to record videos in everyday life and share them through the Internet. Video blog is a recent approach for sharing videos, in which users enjoy expressing themselves in blog posts with attractive videos. Generating such videos, however, requires users to review vast amount of raw videos and edit them appropriately, which keeps users away from doing so. In this paper, we propose a novel video summarization method for helping users to create a video blog post. Unlike typical video summarization methods, the proposed method utilizes the text, which is written for a video blog post, and makes the video summary consistent with the content of the text. For this, we perform video summarization by solving an optimization problem, in which an objective function involves the content similarity between the summarized video and the text. Our user study with 20 participants has demonstrated that our proposed method is suitable to create video blog posts compared with conventional methods for video summarization.

When observing a person (an actor) performing or demonstrating some activity for the purpose of learning the action, it is best for the viewers to be present at the same time and place as the actor. Otherwise, a video must be recorded. However, conventional video only provides two-dimensional (2D) motion, which lacks the original third dimension of motion. In the presence of some ambiguity, it may be hard for the viewer to comprehend the action with only two dimensions, making it harder to learn the action. This paper proposes an augmented reality system to reenact such actions at any time the viewer wants, in order to aid comprehension of 3D motion. In the proposed system, a user first captures the actor's motion and appearance, using a single RGB-D camera. Upon a viewer's request, our system displays the motion from an arbitrary viewpoint using a rough 3D model of the subject, made up of cylinders, and selecting the most appropriate textures based on the viewpoint and the subject's pose. We evaluate the usefulness of the system and the quality of the displayed images by user study.

Detecting important regions in videos has been extensively studied for past decades for their wide variety of applications including video summarization and retargeting. Visual attention models draw much attention for this purpose, which find visually salient regions. However, visual attention models ignore intentionally captured regions (ICRs) derived from videographers' intentions, i.e., what the videographers wanted to capture in their videos. This paper proposes a Markov random field-based ICR model for finding them. Observing that a videographer's intention is embedded into camera motion together with objects' motion, our ICR model uses point trajectory-based features to distinguish ICRs from non-ICRs. It also leverages spatial and temporal consistency of ICRs to improve the performance. We have experimentally demonstrated our ICR model's performance and the difference between ICRs and visually salient regions.

Most social videos, which are uploaded and shared through social networking services (SNSs), e.g., YouTube and Facebook, contain not only intentionally-captured persons (ICPs) but also non-ICPs who are unexpectedly framed in, such as passers-by. Sharing such social videos may infringe on the non-ICPs' privacy but not on the ICPs' in many cases; however, existing systems for video privacy protection simply obscure persons without distinguishing ICPs from non-ICPs. This naive obscuration may spoil the videos. Since this is a critical problem especially for social videos, in this paper, we propose a novel system for automatically generating privacy-protected videos in real-time. Our system localizes ICPs and non-ICPs using ICP detection leveraging the spatial and temporal consistency of ICPs/non-ICPs and obscures the non-ICPs. We have experimentally evaluated the performance of ICP detection and demonstrated the applicability of our system.

Augmented reality (AR) images with virtualized real objects can be used for various applications. However, such AR image generation requires hand-crafted 3D models of that objects, which are usually not available. This paper proposes a view-dependent texture (VDT)- and view-dependent geometry (VDG)-based method for generating high quality AR images, which uses 3D models automatically reconstructed from multiple images. Since the quality of reconstructed 3D models is usually insufficient, the proposed method inflates the objects in the depth map as VDG to repair chipped object boundaries and assigns a color to each pixel based on VDT to reproduce the detail of the objects. Background pixel exposure due to inflation is suppressed by the use of the foreground region extracted from the input images. Our experimental results have demonstrated that the proposed method can successfully reduce above visual artifacts.

Most videos taken by videographers contain intentionally-captured persons (ICPs), who are essential for what the videographers want to express in their video. This paper presents a method to detect ICPs in real-time. Whether a person in a video is an ICP or not is reflected in features such as the person's motion and camera motion, which are thus beneficial for detecting ICPs. However, estimating camera motion is computationally expensive. For real-time detection, we use samples of acceleration and angular velocity obtained from inertial sensors instead of estimating camera motion. Considering that pairwise constraints based on differences between persons' sizes also improve the detection performance, we model the ICPs using Markov random field. We experimentally evaluate the performance of our method and demonstrate that it works in real-time.

Recently, video sharing services such as YouTube and Daily-motion have become popular and many videos taken with mobile video cameras are uploaded to such a video sharing service. However, such videos can infringe on the privacy right of people in the videos because they may contain privacy sensitive information (PSI) of the people, i.e., their appearances. This strongly motivates us to develop a technique to generate privacy-protected videos. In this paper, we propose a novel system for automatic generation of privacy-protected videos based on background estimation. In most conventional techniques, objects that contain PSI are detected and obscured by, e.g., blurring. Conversely, in our system, background pixels are estimated and then substituted with intended human objects that are essential for the camera person's capture intention. We quantitatively evaluate our system to demonstrate its potential applicability.

When camera persons take videos with mobile video cameras, they usually have capture intentions, i.e., what they want to express in their videos, and there are intentionally captured regions (ICRs) in the video frames that are essential for the capture intentions. Extracting ICRs is thus beneficial for wide range of applications such as video summarization and video adaptation for small displays. In this paper, we present a novel method for automatically extracting ICRs. A camera person usually moves his/her camera so that ICRs can be arranged in appropriate positions in video frames; therefore, ICRs can yield specific motion. This observation indicates that such specific motion is a vital cue for extracting ICRs. The proposed method represents motion by point trajectories, which are long-term trajectories of spatially dense points in video frames, and extracts ICRs using an ICR model based on the point trajectories. We experimentally evaluate the proposed method to demonstrate its potential applicability.

In this paper, we propose a method for estimating the user position where a user is holding a microphone in an indoor environment using digital watermarking for audio signals. The proposed method utilizes detection strengths, which are calculated while detecting spread-spectrum-based watermarks. Taking into account delays and attenuation of the watermarked signals emitted from multiple loudspeakers and other factors, we construct a model of detection strengths. The user position is estimated in real-time using the model. The experimental results indicate that the user positions are estimated with 1.3 m of root mean squared error on average for the case where the user is static. We demonstrate that the proposed method successfully estimates the user position even when the user moves.

In a consumer video, there are not only intended objects, which are intentionally captured by the camcorder user, but also unintended objects, which are accidentally framed-in. Since the intended objects are essential to present what the camcorder user wants to express in the video, discriminating the intended objects from the unintended objects are beneficial for many applications, e.g., video summarization, privacy protection, and so forth. In this paper, focusing on human objects, we propose a method for discriminating the intended human objects from the unintended human objects. We evaluated the proposed method using 10 videos captured by 3 camcorder users. The results demonstrate that the proposed method successfully discriminates the intended human objects with 0.45 of recall and 0.80 of precision.

The growing popularity of video sharing services such as YouTube enables us to upload and share consumer generated videos (CGVs) easily, resulting in disclosure of the privacy sensitive information (PSI) of persons, i.e., their appearances. Therefore, we need a technique for automatically protecting the privacy in CGVs; however, the main problem is how to determine PSI regions automatically. In this paper, we propose a novel system for automatically protecting the privacy in CGVs. The proposed system tackles the problem of determining PSI regions by using an intended human object detector that detects human objects which the camera person wanted to capture to achieve his/her capture intention. In addition, the proposed system adopts several PSI obscuring methods such as blocking out, blurring and seam carving. We present the results of subjective evaluations of a privacy protected video in terms of the visual quality and acceptability of PSI disclosure, as well as the performance of the intended human object detector.

In recent years, the problem of camcorder piracy in theaters has become more serious due to technical advances in camcorders. In this paper, as a new deterrent to camcorder piracy, we propose a system for estimating the recording position from which a camcorder recording is made. The system is based on spread-spectrum audio watermarking for the multichannel movie soundtrack. It utilizes a stochastic model of the detection strength, which is calculated in the watermark detection process. Our experimental results show that the system estimates recording positions in an actual theater with a mean estimation error of 0.44 m. The results of our MUSHRA subjective listening tests show the method does not significantly spoil the subjective acoustic quality of the soundtrack. These results indicate that the proposed system is applicable for practical uses.

In this paper, we propose a novel application of digital watermarking, determination of recording locations. This application enables us to determine the seat location in an auditorium where a recording was made. Precisely measured synchronization positions of the spread-spectrum watermarks are used for the determination. To avoid use of mismeasured synchronization positions, the algorithm discards synchronization positions with the corresponding normalized correlation values below a threshold. The experiments with our implementation resulted in accurate determinations; almost all of the locations can be determined within the error of 0.5 m. These experimental results successfully show the potential applicability of our application.