Hitoshi Iyatomi

Quan Huu Cap, Atsushi Fukuda, Hitoshi Iyatomi

Medical images are extremely valuable for supporting medical diagnoses. However, in practice, low-quality (LQ) medical images, such as images that are hazy/blurry, have uneven illumination, or are out of focus, among others, are often obtained during data acquisition. This leads to difficulties in the screening and diagnosis of medical diseases. Several generative adversarial networks (GAN)-based image enhancement methods have been proposed and have shown promising results. However, there is a quality-originality trade-off among these methods in the sense that they produce visually pleasing results but lose the ability to preserve originality, especially the structural inputs. Moreover, to our knowledge, there is no objective metric in evaluating the structure preservation of medical image enhancement methods in unsupervised settings due to the unavailability of paired ground-truth data. In this study, we propose a framework for practical unsupervised medical image enhancement that includes (1) a non-reference objective evaluation of structure preservation for medical image enhancement tasks called Laplacian structural similarity index measure (LaSSIM), which is based on SSIM and the Laplacian pyramid, and (2) a novel unsupervised GAN-based method called Laplacian medical image enhancement (LaMEGAN) to support the improvement of both originality and quality from LQ images. The LaSSIM metric does not require clean reference images and has been shown to be superior to SSIM in capturing image structural changes under image degradations, such as strong blurring on different datasets. The experiments demonstrated that our LaMEGAN achieves a satisfactory balance between quality and originality, with robust structure preservation performance while generating compelling visual results with very high image quality scores. The code will be made available at https://github.com/AillisInc/USPMIE.

Kei Nishimaki, Kumpei Ikuta, Yuto Onga et al.

Content-based image retrieval (CBIR) systems are an emerging technology that supports reading and interpreting medical images. Since 3D brain MR images are high dimensional, dimensionality reduction is necessary for CBIR using machine learning techniques. In addition, for a reliable CBIR system, each dimension in the resulting low-dimensional representation must be associated with a neurologically interpretable region. We propose a localized variational autoencoder (Loc-VAE) that provides neuroanatomically interpretable low-dimensional representation from 3D brain MR images for clinical CBIR. Loc-VAE is based on $β$-VAE with the additional constraint that each dimension of the low-dimensional representation corresponds to a local region of the brain. The proposed Loc-VAE is capable of acquiring representation that preserves disease features and is highly localized, even under high-dimensional compression ratios (4096:1). The low-dimensional representation obtained by Loc-VAE improved the locality measure of each dimension by 4.61 points compared to naive $β$-VAE, while maintaining comparable brain reconstruction capability and information about the diagnosis of Alzheimer's disease.

Ryosuke Wayama, Yuki Sasaki, Satoshi Kagiwada et al.

The development of practical and robust automated diagnostic systems for identifying plant pests is crucial for efficient agricultural production. In this paper, we first investigate three key research questions (RQs) that have not been addressed thus far in the field of image-based plant pest identification. Based on the knowledge gained, we then develop an accurate, robust, and fast plant pest identification framework using 334K images comprising 78 combinations of four plant portions (the leaf front, leaf back, fruit, and flower of cucumber, tomato, strawberry, and eggplant) and 20 pest species captured at 27 farms. The results reveal the following. (1) For an appropriate evaluation of the model, the test data should not include images of the field from which the training images were collected, or other considerations to increase the diversity of the test set should be taken into account. (2) Pre-extraction of ROIs, such as leaves and fruits, helps to improve identification accuracy. (3) Integration of closely related species using the same control methods and cross-crop training methods for the same pests, are effective. Our two-stage plant pest identification framework, enabling ROI detection and convolutional neural network (CNN)-based identification, achieved a highly practical performance of 91.0% and 88.5% in mean accuracy and macro F1 score, respectively, for 12,223 instances of test data of 21 classes collected from unseen fields, where 25 classes of images from 318,971 samples were used for training; the average identification time was 476 ms/image.

Quan Huu Cap, Atsushi Fukuda, Satoshi Kagiwada et al.

With rich annotation information, object detection-based automated plant disease diagnosis systems (e.g., YOLO-based systems) often provide advantages over classification-based systems (e.g., EfficientNet-based), such as the ability to detect disease locations and superior classification performance. One drawback of these detection systems is dealing with unannotated healthy data with no real symptoms present. In practice, healthy plant data appear to be very similar to many disease data. Thus, those models often produce mis-detected boxes on healthy images. In addition, labeling new data for detection models is typically time-consuming. Hard-sample mining (HSM) is a common technique for re-training a model by using the mis-detected boxes as new training samples. However, blindly selecting an arbitrary amount of hard-sample for re-training will result in the degradation of diagnostic performance for other diseases due to the high similarity between disease and healthy data. In this paper, we propose a simple but effective training strategy called hard-sample re-mining (HSReM), which is designed to enhance the diagnostic performance of healthy data and simultaneously improve the performance of disease data by strategically selecting hard-sample training images at an appropriate level. Experiments based on two practical in-field eight-class cucumber and ten-class tomato datasets (42.7K and 35.6K images) show that our HSReM training strategy leads to a substantial improvement in the overall diagnostic performance on large-scale unseen data. Specifically, the object detection model trained using the HSReM strategy not only achieved superior results as compared to the classification-based state-of-the-art EfficientNetV2-Large model and the original object detection model, but also outperformed the model using the HSM strategy in multiple evaluation metrics.

Kohei Iwano, Shogo Shibuya, Satoshi Kagiwada et al.

Recently, object detection methods (OD; e.g., YOLO-based models) have been widely utilized in plant disease diagnosis. These methods demonstrate robustness to distance variations and excel at detecting small lesions compared to classification methods (CL; e.g., CNN models). However, there are issues such as low diagnostic performance for hard-to-detect diseases and high labeling costs. Additionally, since healthy cases cannot be explicitly trained, there is a risk of false positives. We propose the Hierarchical object detection and recognition framework (HODRF), a sophisticated and highly integrated two-stage system that combines the strengths of both OD and CL for plant disease diagnosis. In the first stage, HODRF uses OD to identify regions of interest (ROIs) without specifying the disease. In the second stage, CL diagnoses diseases surrounding the ROIs. HODRF offers several advantages: (1) Since OD detects only one type of ROI, HODRF can detect diseases with limited training images by leveraging its ability to identify other lesions. (2) While OD over-detects healthy cases, HODRF significantly reduces these errors by using CL in the second stage. (3) CL's accuracy improves in HODRF as it identifies diagnostic targets given as ROIs, making it less vulnerable to size changes. (4) HODRF benefits from CL's lower annotation costs, allowing it to learn from a larger number of images. We implemented HODRF using YOLOv7 for OD and EfficientNetV2 for CL and evaluated its performance on a large-scale dataset (4 crops, 20 diseased and healthy classes, 281K images). HODRF outperformed YOLOv7 alone by 5.8 to 21.5 points on healthy data and 0.6 to 7.5 points on macro F1 scores, and it improved macro F1 by 1.1 to 7.2 points over EfficientNetV2.

Shunsuke Kitada, Yuki Iwazaki, Riku Togashi et al.

There is increasing interest in the use of multimodal data in various web applications, such as digital advertising and e-commerce. Typical methods for extracting important information from multimodal data rely on a mid-fusion architecture that combines the feature representations from multiple encoders. However, as the number of modalities increases, several potential problems with the mid-fusion model structure arise, such as an increase in the dimensionality of the concatenated multimodal features and missing modalities. To address these problems, we propose a new concept that considers multimodal inputs as a set of sequences, namely, deep multimodal sequence sets (DM$^2$S$^2$). Our set-aware concept consists of three components that capture the relationships among multiple modalities: (a) a BERT-based encoder to handle the inter- and intra-order of elements in the sequences, (b) intra-modality residual attention (IntraMRA) to capture the importance of the elements in a modality, and (c) inter-modality residual attention (InterMRA) to enhance the importance of elements with modality-level granularity further. Our concept exhibits performance that is comparable to or better than the previous set-aware models. Furthermore, we demonstrate that the visualization of the learned InterMRA and IntraMRA weights can provide an interpretation of the prediction results.

Tsubasa Nakagawa, Shunsuke Kitada, Hitoshi Iyatomi

It is often difficult to correctly infer a writer's emotion from text exchanged online, and differences in recognition between writers and readers can be problematic. In this paper, we propose a new framework for detecting sentences that create differences in emotion recognition between the writer and the reader and for detecting the kinds of expressions that cause such differences. The proposed framework consists of a bidirectional encoder representations from transformers (BERT)-based detector that detects sentences causing differences in emotion recognition and an analysis that acquires expressions that characteristically appear in such sentences. The detector, based on a Japanese SNS-document dataset with emotion labels annotated by both the writer and three readers of the social networking service (SNS) documents, detected "hidden-anger sentences" with AUC = 0.772; these sentences gave rise to differences in the recognition of anger. Because SNS documents contain many sentences whose meaning is extremely difficult to interpret, by analyzing the sentences detected by this detector, we obtained several expressions that appear characteristically in hidden-anger sentences. The detected sentences and expressions do not convey anger explicitly, and it is difficult to infer the writer's anger, but if the implicit anger is pointed out, it becomes possible to guess why the writer is angry. Put into practical use, this framework would likely have the ability to mitigate problems based on misunderstandings.

Shunsuke Kitada, Hitoshi Iyatomi, Yoshifumi Seki

Discontinuing ad creatives at an appropriate time is one of the most important ad operations that can have a significant impact on sales. Such operational support for ineffective ads has been less explored than that for effective ads. After pre-analyzing 1,000,000 real-world ad creatives, we found that there are two types of discontinuation: short-term (i.e., cut-out) and long-term (i.e., wear-out). In this paper, we propose a practical prediction framework for the discontinuation of ad creatives with a hazard function-based loss function inspired by survival analysis. Our framework predicts the discontinuations with a multi-modal deep neural network that takes as input the ad creative (e.g., text, categorical, image, numerical features). To improve the prediction performance for the two different types of discontinuations and for the ad creatives that contribute to sales, we introduce two new techniques: (1) a two-term estimation technique with multi-task learning and (2) a click-through rate-weighting technique for the loss function. We evaluated our framework using the large-scale ad creative dataset, including 10 billion scale impressions. In terms of the concordance index (short: 0.896, long: 0.939, and overall: 0.792), our framework achieved significantly better performance than the conventional method (0.531). Additionally, we confirmed that our framework (i) demonstrated the same degree of discontinuation effect as manual operations for short-term cases, and (ii) accurately predicted the ad discontinuation order, which is important for long-running ad creatives for long-term cases.

Kazuya Ohata, Shunsuke Kitada, Hitoshi Iyatomi

We propose a simple yet effective image captioning framework that can determine the quality of an image and notify the user of the reasons for any flaws in the image. Our framework first determines the quality of images and then generates captions using only those images that are determined to be of high quality. The user is notified by the flaws feature to retake if image quality is low, and this cycle is repeated until the input image is deemed to be of high quality. As a component of the framework, we trained and evaluated a low-quality image detection model that simultaneously learns difficulty in recognizing images and individual flaws, and we demonstrated that our proposal can explain the reasons for flaws with a sufficient score. We also evaluated a dataset with low-quality images removed by our framework and found improved values for all four common metrics (e.g., BLEU-4, METEOR, ROUGE-L, CIDEr), confirming an improvement in general-purpose image captioning capability. Our framework would assist the visually impaired, who have difficulty judging image quality.

Takumi Ohashi, Hitoshi Iyatomi

Large language models (LLMs) are increasingly deployed in multicultural settings; however, systematic evaluation of cultural specificity at the sentence level remains underexplored. We propose the Conceptual Cultural Index (CCI), which estimates cultural specificity at the sentence level. CCI is defined as the difference between the generality estimate within the target culture and the average generality estimate across other cultures. This formulation enables users to operationally control the scope of culture via comparison settings and provides interpretability, since the score derives from the underlying generality estimates. We validate CCI on 400 sentences (200 culture-specific and 200 general), and the resulting score distribution exhibits the anticipated pattern: higher for culture-specific sentences and lower for general ones. For binary separability, CCI outperforms direct LLM scoring, yielding more than a 10-point improvement in AUC for models specialized to the target culture. Our code is available at https://github.com/IyatomiLab/CCI .

Michito Takeshita, Takuro Kawada, Takumi Ohashi et al.

AI agents that interact with graphical user interfaces (GUIs) require effective observation representations for reliable grounding. The accessibility tree is a commonly used text-based format that encodes UI element attributes, but it suffers from redundancy and lacks structural information such as spatial relationships among elements. We propose A11y-Compressor, a framework that transforms linearized accessibility trees into compact and structured representations. Our implementation, Compressed-a11y, applies a lightweight and structured transformation pipeline with modal detection, redundancy reduction, and semantic structuring. Experiments on the OSWorld benchmark show that Compressed-a11y reduces input tokens to 22% of the original while improving task success rates by 5.1 percentage points on average.

Takumi Aoki, Shunsuke Kitada, Hitoshi Iyatomi

We propose a new character-based text classification framework for non-alphabetic languages, such as Chinese and Japanese. Our framework consists of a variational character encoder (VCE) and character-level text classifier. The VCE is composed of a $β$-variational auto-encoder ($β$-VAE) that learns the proposed glyph-aware disentangled character embedding (GDCE). Since our GDCE provides zero-mean unit-variance character embeddings that are dimensionally independent, it is applicable for our interpretable data augmentation, namely, semantic sub-character augmentation (SSA). In this paper, we evaluated our framework using Japanese text classification tasks at the document- and sentence-level. We confirmed that our GDCE and SSA not only provided embedding interpretability but also improved the classification performance. Our proposal achieved a competitive result to the state-of-the-art model while also providing model interpretability. Our code is available on https://github.com/IyatomiLab/GDCE-SSA

Shunsuke Kitada, Hitoshi Iyatomi

Although attention mechanisms have been applied to a variety of deep learning models and have been shown to improve the prediction performance, it has been reported to be vulnerable to perturbations to the mechanism. To overcome the vulnerability to perturbations in the mechanism, we are inspired by adversarial training (AT), which is a powerful regularization technique for enhancing the robustness of the models. In this paper, we propose a general training technique for natural language processing tasks, including AT for attention (Attention AT) and more interpretable AT for attention (Attention iAT). The proposed techniques improved the prediction performance and the model interpretability by exploiting the mechanisms with AT. In particular, Attention iAT boosts those advantages by introducing adversarial perturbation, which enhances the difference in the attention of the sentences. Evaluation experiments with ten open datasets revealed that AT for attention mechanisms, especially Attention iAT, demonstrated (1) the best performance in nine out of ten tasks and (2) more interpretable attention (i.e., the resulting attention correlated more strongly with gradient-based word importance) for all tasks. Additionally, the proposed techniques are (3) much less dependent on perturbation size in AT. Our code is available at https://github.com/shunk031/attention-meets-perturbation

Quan Huu Cap, Hiroyuki Uga, Satoshi Kagiwada et al.

Many applications for the automated diagnosis of plant disease have been developed based on the success of deep learning techniques. However, these applications often suffer from overfitting, and the diagnostic performance is drastically decreased when used on test datasets from new environments. In this paper, we propose LeafGAN, a novel image-to-image translation system with own attention mechanism. LeafGAN generates a wide variety of diseased images via transformation from healthy images, as a data augmentation tool for improving the performance of plant disease diagnosis. Thanks to its own attention mechanism, our model can transform only relevant areas from images with a variety of backgrounds, thus enriching the versatility of the training images. Experiments with five-class cucumber disease classification show that data augmentation with vanilla CycleGAN cannot help to improve the generalization, i.e., disease diagnostic performance increased by only 0.7% from the baseline. In contrast, LeafGAN boosted the diagnostic performance by 7.4%. We also visually confirmed the generated images by our LeafGAN were much better quality and more convincing than those generated by vanilla CycleGAN. The code is available publicly at: https://github.com/IyatomiLab/LeafGAN.

Daichi Nagai, Ryugo Morita, Shunsuke Kitada et al.

Despite the remarkable success of text-to-image diffusion models, their output of a single, flattened image remains a critical bottleneck for professional applications requiring layer-wise control. Existing solutions either rely on fine-tuning with large, inaccessible datasets or are training-free yet limited to generating isolated foreground elements, failing to produce a complete and coherent scene. To address this, we introduce the Training-free Noise Transplantation and Cultivation Diffusion Model (TAUE), a novel framework for zero-shot, layer-wise image generation. Our core technique, Noise Transplantation and Cultivation (NTC), extracts intermediate latent representations from both foreground and composite generation processes, transplanting them into the initial noise for subsequent layers. This ensures semantic and structural coherence across foreground, background, and composite layers, enabling consistent, multi-layered outputs without requiring fine-tuning or auxiliary datasets. Extensive experiments show that our training-free method achieves performance comparable to fine-tuned methods, enhancing layer-wise consistency while maintaining high image quality and fidelity. TAUE not only eliminates costly training and dataset requirements but also unlocks novel downstream applications, such as complex compositional editing, paving the way for more accessible and controllable generative workflows.

Shuhei Tomoshige, Hayato Muraki, Kenichi Oishi et al.

Current methods for searching brain MR images rely on text-based approaches, highlighting a significant need for content-based image retrieval (CBIR) systems. Directly applying 3D brain MR images to machine learning models offers the benefit of effectively learning the brain's structure; however, building the generalized model necessitates a large amount of training data. While models that consider depth direction and utilize continuous 2D slices have demonstrated success in segmentation and classification tasks involving 3D data, concerns remain. Specifically, using general 2D slices may lead to the oversight of pathological features and discontinuities in depth direction information. Furthermore, to the best of the authors' knowledge, there have been no attempts to develop a practical CBIR system that preserves the entire brain's structural information. In this study, we propose an interpretable CBIR method for brain MR images, named iCBIR-Sli (Interpretable CBIR with 2D Slice Embedding), which, for the first time globally, utilizes a series of 2D slices. iCBIR-Sli addresses the challenges associated with using 2D slices by effectively aggregating slice information, thereby achieving low-dimensional representations with high completeness, usability, robustness, and interoperability, which are qualities essential for effective CBIR. In retrieval evaluation experiments utilizing five publicly available brain MR datasets (ADNI2/3, OASIS3/4, AIBL) for Alzheimer's disease and cognitively normal, iCBIR-Sli demonstrated top-1 retrieval performance (macro F1 = 0.859), comparable to existing deep learning models explicitly designed for classification, without the need for an external classifier. Additionally, the method provided high interpretability by clearly identifying the brain regions indicative of the searched-for disease.

Yuji Arima, Satoshi Kagiwada, Hitoshi Iyatomi

Recent studies on plant disease diagnosis using machine learning (ML) have highlighted concerns about the overestimated diagnostic performance due to inappropriate data partitioning, where training and test datasets are derived from the same source (domain). Plant disease diagnosis presents a challenging classification task, characterized by its fine-grained nature, vague symptoms, and the extensive variability of image features within each domain. In this study, we propose the concept of Discriminative Difficulty Distance (DDD), a novel metric designed to quantify the domain gap between training and test datasets while assessing the classification difficulty of test data. DDD provides a valuable tool for identifying insufficient diversity in training data, thus supporting the development of more diverse and robust datasets. We investigated multiple image encoders trained on different datasets and examined whether the distances between datasets, measured using low-dimensional representations generated by the encoders, are suitable as a DDD metric. The study utilized 244,063 plant disease images spanning four crops and 34 disease classes collected from 27 domains. As a result, we demonstrated that even if the test images are from different crops or diseases than those used to train the encoder, incorporating them allows the construction of a distance measure for a dataset that strongly correlates with the difficulty of diagnosis indicated by the disease classifier developed independently. Compared to the base encoder, pre-trained only on ImageNet21K, the correlation higher by 0.106 to 0.485, reaching a maximum of 0.909.

Shoma Kudo, Satoshi Kagiwada, Hitoshi Iyatomi

Numerous studies have explored image-based automated systems for plant disease diagnosis, demonstrating impressive diagnostic capabilities. However, recent large-scale analyses have revealed a critical limitation: that the diagnostic capability suffers significantly when validated on images captured in environments (domains) differing from those used during training. This shortfall stems from the inherently limited dataset size and the diverse manifestation of disease symptoms, combined with substantial variations in cultivation environments and imaging conditions, such as equipment and composition. These factors lead to insufficient variety in training data, ultimately constraining the system's robustness and generalization. To address these challenges, we propose Few-shot Metric Domain Adaptation (FMDA), a flexible and effective approach for enhancing diagnostic accuracy in practical systems, even when only limited target data is available. FMDA reduces domain discrepancies by introducing a constraint to the diagnostic model that minimizes the "distance" between feature spaces of source (training) data and target data with limited samples. FMDA is computationally efficient, requiring only basic feature distance calculations and backpropagation, and can be seamlessly integrated into any machine learning (ML) pipeline. In large-scale experiments, involving 223,015 leaf images across 20 fields and 3 crop species, FMDA achieved F1 score improvements of 11.1 to 29.3 points compared to cases without target data, using only 10 images per disease from the target domain. Moreover, FMDA consistently outperformed fine-tuning methods utilizing the same data, with an average improvement of 8.5 points.

Takuro Kawada, Shunsuke Kitada, Sota Nemoto et al.

Graphical Abstracts (GAs) play a crucial role in visually conveying the key findings of scientific papers. While recent research has increasingly incorporated visual materials such as Figure 1 as de facto GAs, their potential to enhance scientific communication remains largely unexplored. Moreover, designing effective GAs requires advanced visualization skills, creating a barrier to their widespread adoption. To tackle these challenges, we introduce SciGA-145k, a large-scale dataset comprising approximately 145,000 scientific papers and 1.14 million figures, explicitly designed for supporting GA selection and recommendation as well as facilitating research in automated GA generation. As a preliminary step toward GA design support, we define two tasks: 1) Intra-GA recommendation, which identifies figures within a given paper that are well-suited to serve as GAs, and 2) Inter-GA recommendation, which retrieves GAs from other papers to inspire the creation of new GAs. We provide reasonable baseline models for these tasks. Furthermore, we propose Confidence Adjusted top-1 ground truth Ratio (CAR), a novel recommendation metric that offers a fine-grained analysis of model behavior. CAR addresses limitations in traditional ranking-based metrics by considering cases where multiple figures within a paper, beyond the explicitly labeled GA, may also serve as GAs. By unifying these tasks and metrics, our SciGA-145k establishes a foundation for advancing visual scientific communication while contributing to the development of AI for Science.

Shuya Tobari, Shuhei Tomoshige, Hayato Muraki et al.

When conducting large-scale studies that collect brain MR images from multiple facilities, the impact of differences in imaging equipment and protocols at each site cannot be ignored, and this domain gap has become a significant issue in recent years. In this study, we propose a new low-dimensional representation (LDR) acquisition method called style encoder adversarial domain adaptation (SE-ADA) to realize content-based image retrieval (CBIR) of brain MR images. SE-ADA reduces domain differences while preserving pathological features by separating domain-specific information from LDR and minimizing domain differences using adversarial learning. In evaluation experiments comparing SE-ADA with recent domain harmonization methods on eight public brain MR datasets (ADNI1/2/3, OASIS1/2/3/4, PPMI), SE-ADA effectively removed domain information while preserving key aspects of the original brain structure and demonstrated the highest disease search accuracy.

Keima Abe, Hayato Muraki, Shuhei Tomoshige et al.

Medical images like MR scans often show domain shifts across imaging sites due to scanner and protocol differences, which degrade machine learning performance in tasks such as disease classification. Domain harmonization is thus a critical research focus. Recent approaches encode brain images $\boldsymbol{x}$ into a low-dimensional latent space $\boldsymbol{z}$, then disentangle it into $\boldsymbol{z_u}$ (domain-invariant) and $\boldsymbol{z_d}$ (domain-specific), achieving strong results. However, these methods often lack interpretability$-$an essential requirement in medical applications$-$leaving practical issues unresolved. We propose Pseudo-Linear-Style Encoder Adversarial Domain Adaptation (PL-SE-ADA), a general framework for domain harmonization and interpretable representation learning that preserves disease-relevant information in brain MR images. PL-SE-ADA includes two encoders $f_E$ and $f_{SE}$ to extract $\boldsymbol{z_u}$ and $\boldsymbol{z_d}$, a decoder to reconstruct the image $f_D$, and a domain predictor $g_D$. Beyond adversarial training between the encoder and domain predictor, the model learns to reconstruct the input image $\boldsymbol{x}$ by summing reconstructions from $\boldsymbol{z_u}$ and $\boldsymbol{z_d}$, ensuring both harmonization and informativeness. Compared to prior methods, PL-SE-ADA achieves equal or better performance in image reconstruction, disease classification, and domain recognition. It also enables visualization of both domain-independent brain features and domain-specific components, offering high interpretability across the entire framework.

Takafumi Nogami, Satoshi Kagiwada, Hitoshi Iyatomi

Various deep learning-based systems have been proposed for accurate and convenient plant disease diagnosis, achieving impressive performance. However, recent studies show that these systems often fail to maintain diagnostic accuracy on images captured under different conditions from the training environment -- an essential criterion for model robustness. Many deep learning methods have shown high accuracy in plant disease diagnosis. However, they often struggle to generalize to images taken in conditions that differ from the training setting. This drop in performance stems from the subtle variability of disease symptoms and domain gaps -- differences in image context and environment. The root cause is the limited diversity of training data relative to task complexity, making even advanced models vulnerable in unseen domains. To tackle this challenge, we propose a simple yet highly adaptable learning framework called Target-Aware Metric Learning with Prioritized Sampling (TMPS), grounded in metric learning. TMPS operates under the assumption of access to a limited number of labeled samples from the target (deployment) domain and leverages these samples effectively to improve diagnostic robustness. We assess TMPS on a large-scale automated plant disease diagnostic task using a dataset comprising 223,073 leaf images sourced from 23 agricultural fields, spanning 21 diseases and healthy instances across three crop species. By incorporating just 10 target domain samples per disease into training, TMPS surpasses models trained using the same combined source and target samples, and those fine-tuned with these target samples after pre-training on source data. It achieves average macro F1 score improvements of 7.3 and 3.6 points, respectively, and a remarkable 18.7 and 17.1 point improvement over the baseline and conventional metric learning.

Takumi Ohashi, Tsubasa Nakagawa, Hitoshi Iyatomi

Rapid advancements in artificial intelligence (AI) have made it crucial to integrate moral reasoning into AI systems. However, existing models and datasets often overlook regional and cultural differences. To address this shortcoming, we have expanded the JCommonsenseMorality (JCM) dataset, the only publicly available dataset focused on Japanese morality. The Extended JCM (eJCM) has grown from the original 13,975 sentences to 31,184 sentences using our proposed sentence expansion method called Masked Token and Label Enhancement (MTLE). MTLE selectively masks important parts of sentences related to moral judgment and replaces them with alternative expressions generated by a large language model (LLM), while re-assigning appropriate labels. The model trained using our eJCM achieved an F1 score of 0.857, higher than the scores for the original JCM (0.837), ChatGPT one-shot classification (0.841), and data augmented using AugGPT, a state-of-the-art augmentation method (0.850). Specifically, in complex moral reasoning tasks unique to Japanese culture, the model trained with eJCM showed a significant improvement in performance (increasing from 0.681 to 0.756) and achieved a performance close to that of GPT-4 Turbo (0.787). These results demonstrate the validity of the eJCM dataset and the importance of developing models and datasets that consider the cultural context.

Sota Nemoto, Shunsuke Kitada, Hitoshi Iyatomi

Data imbalance presents a significant challenge in various machine learning (ML) tasks, particularly named entity recognition (NER) within natural language processing (NLP). NER exhibits a data imbalance with a long-tail distribution, featuring numerous minority classes (i.e., entity classes) and a single majority class (i.e., O-class). This imbalance leads to misclassifications of the entity classes as the O-class. To tackle this issue, we propose a simple and effective learning method named majority or minority (MoM) learning. MoM learning incorporates the loss computed only for samples whose ground truth is the majority class into the loss of the conventional ML model. Evaluation experiments on four NER datasets (Japanese and English) showed that MoM learning improves prediction performance of the minority classes without sacrificing the performance of the majority class and is more effective than widely known and state-of-the-art methods. We also evaluated MoM learning using frameworks as sequential labeling and machine reading comprehension, which are commonly used in NER. Furthermore, MoM learning has achieved consistent performance improvements regardless of language or framework.

Hideaki Okamoto, Quan Huu Cap, Takakiyo Nomura et al.

Endoscopy is widely used to diagnose gastric cancer and has a high diagnostic performance, but it must be performed by a physician, which limits the number of people who can be diagnosed. In contrast, gastric X-rays can be taken by radiographers, thus allowing a much larger number of patients to undergo imaging. However, the diagnosis of X-ray images relies heavily on the expertise and experience of physicians, and few machine learning methods have been developed to assist in this process. We propose a novel and practical gastric cancer diagnostic support system for gastric X-ray images that will enable more people to be screened. The system is based on a general deep learning-based object detection model and incorporates two novel techniques: refined probabilistic stomach image augmentation (R-sGAIA) and hard boundary box training (HBBT). R-sGAIA enhances the probabilistic gastric fold region and provides more learning patterns for cancer detection models. HBBT is an efficient training method that improves model performance by allowing the use of unannotated negative (i.e., healthy control) samples, which are typically unusable in conventional detection models. The proposed system achieved a sensitivity (SE) for gastric cancer of 90.2\%, higher than that of an expert (85.5%). Under these conditions, two out of five candidate boxes identified by the system were cancerous (precision = 42.5%), with an image processing speed of 0.51 seconds per image. The system also outperformed methods using the same object detection model and state-of-the-art data augmentation by showing a 5.9-point improvement in the F1 score. In summary, this system efficiently identifies areas for radiologists to examine within a practical time frame, thus significantly reducing their workload.

Hayato Arai, Yuto Onga, Kumpei Ikuta et al.

To build a robust and practical content-based image retrieval (CBIR) system that is applicable to a clinical brain MRI database, we propose a new framework -- Disease-oriented image embedding with pseudo-scanner standardization (DI-PSS) -- that consists of two core techniques, data harmonization and a dimension reduction algorithm. Our DI-PSS uses skull stripping and CycleGAN-based image transformations that map to a standard brain followed by transformation into a brain image taken with a given reference scanner. Then, our 3D convolutioinal autoencoders (3D-CAE) with deep metric learning acquires a low-dimensional embedding that better reflects the characteristics of the disease. The effectiveness of our proposed framework was tested on the T1-weighted MRIs selected from the Alzheimer's Disease Neuroimaging Initiative and the Parkinson's Progression Markers Initiative. We confirmed that our PSS greatly reduced the variability of low-dimensional embeddings caused by different scanner and datasets. Compared with the baseline condition, our PSS reduced the variability in the distance from Alzheimer's disease (AD) to clinically normal (CN) and Parkinson disease (PD) cases by 15.8-22.6% and 18.0-29.9%, respectively. These properties allow DI-PSS to generate lower dimensional representations that are more amenable to disease classification. In AD and CN classification experiments based on spectral clustering, PSS improved the average accuracy and macro-F1 by 6.2% and 10.7%, respectively. Given the potential of the DI-PSS for harmonizing images scanned by MRI scanners that were not used to scan the training data, we expect that the DI-PSS is suitable for application to a large number of legacy MRIs scanned in heterogeneous environments.

Shunsuke Kitada, Hitoshi Iyatomi

Although attention mechanisms have become fundamental components of deep learning models, they are vulnerable to perturbations, which may degrade the prediction performance and model interpretability. Adversarial training (AT) for attention mechanisms has successfully reduced such drawbacks by considering adversarial perturbations. However, this technique requires label information, and thus, its use is limited to supervised settings. In this study, we explore the concept of incorporating virtual AT (VAT) into the attention mechanisms, by which adversarial perturbations can be computed even from unlabeled data. To realize this approach, we propose two general training techniques, namely VAT for attention mechanisms (Attention VAT) and "interpretable" VAT for attention mechanisms (Attention iVAT), which extend AT for attention mechanisms to a semi-supervised setting. In particular, Attention iVAT focuses on the differences in attention; thus, it can efficiently learn clearer attention and improve model interpretability, even with unlabeled data. Empirical experiments based on six public datasets revealed that our techniques provide better prediction performance than conventional AT-based as well as VAT-based techniques, and stronger agreement with evidence that is provided by humans in detecting important words in sentences. Moreover, our proposal offers these advantages without needing to add the careful selection of unlabeled data. That is, even if the model using our VAT-based technique is trained on unlabeled data from a source other than the target task, both the prediction performance and model interpretability can be improved.

Kasumi Obi, Quan Huu Cap, Noriko Umegaki-Arao et al.

Although highly accurate automated diagnostic techniques for melanoma have been reported, the realization of a system capable of providing diagnostic evidence based on medical indices remains an open issue because of difficulties in obtaining reliable training data. In this paper, we propose bulk production augmentation (BPA) to generate high-quality, diverse pseudo-skin tumor images with the desired structural malignant features for additional training images from a limited number of labeled images. The proposed BPA acts as an effective data augmentation in constructing the feature detector for the atypical pigment network (APN), which is a key structure in melanoma diagnosis. Experiments show that training with images generated by our BPA largely boosts the APN detection performance by 20.0 percentage points in the area under the receiver operating characteristic curve, which is 11.5 to 13.7 points higher than that of conventional CycleGAN-based augmentations in AUC.

Quan Huu Cap, Hitoshi Iyatomi, Atsushi Fukuda

Medical images have been indispensable and useful tools for supporting medical experts in making diagnostic decisions. However, taken medical images especially throat and endoscopy images are normally hazy, lack of focus, or uneven illumination. Thus, these could difficult the diagnosis process for doctors. In this paper, we propose MIINet, a novel image-to-image translation network for improving quality of medical images by unsupervised translating low-quality images to the high-quality clean version. Our MIINet is not only capable of generating high-resolution clean images, but also preserving the attributes of original images, making the diagnostic more favorable for doctors. Experiments on dehazing 100 practical throat images show that our MIINet largely improves the mean doctor opinion score (MDOS), which assesses the quality and the reproducibility of the images from the baseline of 2.36 to 4.11, while dehazed images by CycleGAN got lower score of 3.83. The MIINet is confirmed by three physicians to be satisfying in supporting throat disease diagnostic from original low-quality images.

Quan Huu Cap, Hiroki Tani, Hiroyuki Uga et al.

The collection of high-resolution training data is crucial in building robust plant disease diagnosis systems, since such data have a significant impact on diagnostic performance. However, they are very difficult to obtain and are not always available in practice. Deep learning-based techniques, and particularly generative adversarial networks (GANs), can be applied to generate high-quality super-resolution images, but these methods often produce unexpected artifacts that can lower the diagnostic performance. In this paper, we propose a novel artifact-suppression super-resolution method that is specifically designed for diagnosing leaf disease, called Leaf Artifact-Suppression Super Resolution (LASSR). Thanks to its own artifact removal module that detects and suppresses artifacts to a considerable extent, LASSR can generate much more pleasing, high-quality images compared to the state-of-the-art ESRGAN model. Experiments based on a five-class cucumber disease (including healthy) discrimination model show that training with data generated by LASSR significantly boosts the performance on an unseen test dataset by nearly 22% compared with the baseline, and that our approach is more than 2% better than a model trained with images generated by ESRGAN.

Mahmoud Daif, Shunsuke Kitada, Hitoshi Iyatomi

Classical and some deep learning techniques for Arabic text classification often depend on complex morphological analysis, word segmentation, and hand-crafted feature engineering. These could be eliminated by using character-level features. We propose a novel end-to-end Arabic document classification framework, Arabic document image-based classifier (AraDIC), inspired by the work on image-based character embeddings. AraDIC consists of an image-based character encoder and a classifier. They are trained in an end-to-end fashion using the class balanced loss to deal with the long-tailed data distribution problem. To evaluate the effectiveness of AraDIC, we created and published two datasets, the Arabic Wikipedia title (AWT) dataset and the Arabic poetry (AraP) dataset. To the best of our knowledge, this is the first image-based character embedding framework addressing the problem of Arabic text classification. We also present the first deep learning-based text classifier widely evaluated on modern standard Arabic, colloquial Arabic and classical Arabic. AraDIC shows performance improvement over classical and deep learning baselines by 12.29% and 23.05% for the micro and macro F-score, respectively.

Yuto Onga, Shingo Fujiyama, Hayato Arai et al.

Increasing numbers of MRI brain scans, improvements in image resolution, and advancements in MRI acquisition technology are causing significant increases in the demand for and burden on radiologists' efforts in terms of reading and interpreting brain MRIs. Content-based image retrieval (CBIR) is an emerging technology for reducing this burden by supporting the reading of medical images. High dimensionality is a major challenge in developing a CBIR system that is applicable for 3D brain MRIs. In this study, we propose a system called disease-oriented data concentration with metric learning (DDCML). In DDCML, we introduce deep metric learning to a 3D convolutional autoencoder (CAE). Our proposed DDCML scheme achieves a high dimensional compression rate (4096:1) while preserving the disease-related anatomical features that are important for medical image classification. The low-dimensional representation obtained by DDCML improved the clustering performance by 29.1\% compared to plain 3D-CAE in terms of discriminating Alzheimer's disease patients from healthy subjects, and successfully reproduced the relationships of the severity of disease categories that were not included in the training.

Quan Huu Cap, Hiroki Tani, Hiroyuki Uga et al.

Automated plant diagnosis using images taken from a distance is often insufficient in resolution and degrades diagnostic accuracy since the important external characteristics of symptoms are lost. In this paper, we first propose an effective pre-processing method for improving the performance of automated plant disease diagnosis systems using super-resolution techniques. We investigate the efficiency of two different super-resolution methods by comparing the disease diagnostic performance on the practical original high-resolution, low-resolution, and super-resolved cucumber images. Our method generates super-resolved images that look very close to natural images with 4$\times$ upscaling factors and is capable of recovering the lost detailed symptoms, largely boosting the diagnostic performance. Our model improves the disease classification accuracy by 26.9% over the bicubic interpolation method of 65.6% and shows a small gap (3% lower) between the original result of 95.5%.

Takumi Saikawa, Quan Huu Cap, Satoshi Kagiwada et al.

In image-based plant diagnosis, clues related to diagnosis are often unclear, and the other factors such as image backgrounds often have a significant impact on the final decision. As a result, overfitting due to latent similarities in the dataset often occurs, and the diagnostic performance on real unseen data (e,g. images from other farms) is usually dropped significantly. However, this problem has not been sufficiently explored, since many systems have shown excellent diagnostic performance due to the bias caused by the similarities in the dataset. In this study, we investigate this problem with experiments using more than 50,000 images of cucumber leaves, and propose an anti-overfitting pretreatment (AOP) for realizing practical image-based plant diagnosis systems. The AOP detects the area of interest (leaf, fruit etc.) and performs brightness calibration as a preprocessing step. The experimental results demonstrate that our AOP can improve the accuracy of diagnosis for unknown test images from different farms by 12.2% in a practical setting.

Katsumasa Suwa, Quan Huu Cap, Ryunosuke Kotani et al.

Practical automated detection and diagnosis of plant disease from wide-angle images (i.e. in-field images containing multiple leaves using a fixed-position camera) is a very important application for large-scale farm management, in view of the need to ensure global food security. However, developing automated systems for disease diagnosis is often difficult, because labeling a reliable wide-angle disease dataset from actual field images is very laborious. In addition, the potential similarities between the training and test data lead to a serious problem of model overfitting. In this paper, we investigate changes in performance when applying disease diagnosis systems to different scenarios involving wide-angle cucumber test data captured on real farms, and propose an effective diagnostic strategy. We show that leading object recognition techniques such as SSD and Faster R-CNN achieve excellent end-to-end disease diagnostic performance only for a test dataset that is collected from the same population as the training dataset (with F1-score of 81.5% - 84.1% for diagnosed cases of disease), but their performance markedly deteriorates for a completely different test dataset (with F1-score of 4.4 - 6.2%). In contrast, our proposed two-stage systems using independent leaf detection and leaf diagnosis stages attain a promising disease diagnostic performance that is more than six times higher than end-to-end systems (with F1-score of 33.4 - 38.9%) on an unseen target dataset. We also confirm the efficiency of our proposal based on visual assessment, concluding that a two-stage model is a suitable and reasonable choice for practical applications.

Shunsuke Kitada, Hitoshi Iyatomi, Yoshifumi Seki

Accurately predicting conversions in advertisements is generally a challenging task, because such conversions do not occur frequently. In this paper, we propose a new framework to support creating high-performing ad creatives, including the accurate prediction of ad creative text conversions before delivering to the consumer. The proposed framework includes three key ideas: multi-task learning, conditional attention, and attention highlighting. Multi-task learning is an idea for improving the prediction accuracy of conversion, which predicts clicks and conversions simultaneously, to solve the difficulty of data imbalance. Furthermore, conditional attention focuses attention of each ad creative with the consideration of its genre and target gender, thus improving conversion prediction accuracy. Attention highlighting visualizes important words and/or phrases based on conditional attention. We evaluated the proposed framework with actual delivery history data (14,000 creatives displayed more than a certain number of times from Gunosy Inc.), and confirmed that these ideas improve the prediction performance of conversions, and visualize noteworthy words according to the creatives' attributes.

Shunsuke Kitada, Ryunosuke Kotani, Hitoshi Iyatomi

For analysing and/or understanding languages having no word boundaries based on morphological analysis such as Japanese, Chinese, and Thai, it is desirable to perform appropriate word segmentation before word embeddings. But it is inherently difficult in these languages. In recent years, various language models based on deep learning have made remarkable progress, and some of these methodologies utilizing character-level features have successfully avoided such a difficult problem. However, when a model is fed character-level features of the above languages, it often causes overfitting due to a large number of character types. In this paper, we propose a CE-CLCNN, character-level convolutional neural networks using a character encoder to tackle these problems. The proposed CE-CLCNN is an end-to-end learning model and has an image-based character encoder, i.e. the CE-CLCNN handles each character in the target document as an image. Through various experiments, we found and confirmed that our CE-CLCNN captured closely embedded features for visually and semantically similar characters and achieves state-of-the-art results on several open document classification tasks. In this paper we report the performance of our CE-CLCNN with the Wikipedia title estimation task and analyse the internal behaviour.

Shunsuke Kitada, Hitoshi Iyatomi

In this report, we introduce the outline of our system in Task 3: Disease Classification of ISIC 2018: Skin Lesion Analysis Towards Melanoma Detection. We fine-tuned multiple pre-trained neural network models based on Squeeze-and-Excitation Networks (SENet) which achieved state-of-the-art results in the field of image recognition. In addition, we used the mean teachers as a semi-supervised learning framework and introduced some specially designed data augmentation strategies for skin lesion analysis. We confirmed our data augmentation strategy improved classification performance and demonstrated 87.2% in balanced accuracy on the official ISIC2018 validation dataset.

M. Emre Celebi, Quan Wen, Sae Hwang et al.

Dermoscopy is one of the major imaging modalities used in the diagnosis of melanoma and other pigmented skin lesions. Due to the difficulty and subjectivity of human interpretation, automated analysis of dermoscopy images has become an important research area. Border detection is often the first step in this analysis. In many cases, the lesion can be roughly separated from the background skin using a thresholding method applied to the blue channel. However, no single thresholding method appears to be robust enough to successfully handle the wide variety of dermoscopy images encountered in clinical practice. In this paper, we present an automated method for detecting lesion borders in dermoscopy images using ensembles of thresholding methods. Experiments on a difficult set of 90 images demonstrate that the proposed method is robust, fast, and accurate when compared to nine state-of-the-art methods.