In this paper, we discuss the initial attempts at boosting understanding human language based on deep-learning models with quantum computing. We successfully train a quantum-enhanced Long Short-Term Memory network to perform parts-of-speech tagging task via numerical simulations. Moreover, Transformer is proposed sentiment analysis existing dataset.

In this work, we propose an AI-based method that intends to improve the conventional retinal disease treatment procedure and help ophthalmologists increase diagnosis efficiency accuracy. The proposed is composed of a deep neural networks-based (DNN-based) module, including identifier clinical description generator, DNN visual explanation module. To train validate effectiveness our DNN-based large-scale image dataset. Also, as ground truth, provide dataset manually labeled by qualitatively...

When video collections become huge, how to explore both within and across videos efficiently is challenging. Video summarization one of the ways tackle this issue. Traditional approaches limit effectiveness exploration because they only generate fixed summary for a given input independent information need user. In work, we introduce method which takes text-based query as generates corresponding it. We do so by modeling supervised learning problem propose an end-to-end deep based...

Low-rank adaptation (LoRA) has been demonstrated effective in reducing the trainable parameter number when fine-tuning a large foundation model (LLM). However, it still encounters computational and memory challenges scaling to larger models or addressing more complex task adaptation. In this work, we introduce Sparse Spectrum Adaptation via Discrete Hartley Transformation (SSH), novel approach that significantly reduces of parameters while enhancing performance. It selects most informative...

Few-shot learning is a nascent research topic, motivated by the fact that traditional deep requires tremendous amounts of data. In this work, we propose new task along direction, call few-shot common-localization. Given few weakly-supervised support images, aim to localize common object in query image without any box annotation. This differs from standard settings, since address localization problem, rather than global classification problem. To tackle network aims get most out and images....

Traditional video summarization methods generate fixed representations regardless of user interest. Therefore such limit users' expectations in content search and exploration scenarios. Multi-modal is one the utilized to address this problem. When multi-modal used help exploration, a text-based query considered as main drivers summary generation, it user-defined. Thus, encoding both effectively important for task summarization. In work, new method proposed that uses specialized attention...

Image search stands as a pivotal task in multimedia and computer vision, finding applications across diverse domains, ranging from internet to medical diagnostics. Conventional image systems operate by accepting textual or visual queries, retrieving the top-relevant candidate results database. However, prevalent methods often rely on single-turn procedures, introducing potential inaccuracies limited recall. These also face challenges, such vocabulary mismatch semantic gap, constraining their...

Deep neural networks have been playing an essential role in many computer vision tasks including Visual Question Answering (VQA). Until recently, the study of their accuracy was main focus research but now there is a trend toward assessing robustness these models against adversarial attacks by evaluating tolerance to varying noise levels. In VQA, can target image and/or proposed question and yet lack proper analysis later. this work, we propose flexible framework that focuses on language...

Automatic machine learning-based (ML-based) medical report generation systems for retinal images suffer from a relative lack of interpretability. Hence, such ML-based are still not widely accepted. The main reason is that trust one the important motivating aspects interpretability and humans do blindly. Precise technical definitions consensus. it difficult to make human-comprehensible system. Heat maps/saliency maps, i.e., post-hoc explanation approaches, used improve systems. However, they...

The goal of video summarization is to automatically shorten videos such that it conveys the overall story without losing relevant information. In many application scenarios, improper can have a large impact. For example in forensics, quality generated summary will affect an investigator's judgment while journalism might yield undesired bias. Because this, modeling explainability key concern. One best ways address challenge uncover causal relations steer process and lead result. Current...

Medical image captioning automatically generates a medical description to describe the content of given image. Traditional models create based on single input only. Hence, an abstract or concept is hard be generated traditional approach. Such method limits effectiveness captioning. Multi-modal one approaches utilized address this problem. In multi-modal captioning, textual input, e.g., expert-defined keywords, considered as main drivers generation. Thus, encoding and effectively are both...

Automatic clinical diagnosis of retinal diseases has emerged as a promising approach to facilitate discovery in areas with limited access specialists. We propose novel visual-assisted hybrid model based on the support vector machine (SVM) and deep neural networks (DNNs). The incorporates complementary strengths DNNs SVM. Furthermore, we present new retina label collection for ophthalmology incorporating 32 classes. Using EyeNet, our achieves 89.73% accuracy performance is comparable...

Automatically generating medical reports for retinal images is one of the promising ways to help ophthalmologists reduce their workload and improve work efficiency. In this work, we propose a new context-driven encoding network automatically generate images. The proposed model mainly composed multi-modal input encoder fused-feature decoder. Our experimental results show that our method capable effectively leveraging interactive information between image context, i.e., keywords in case....

Taking an image and question as the input of our method, it can output text-based answer query about given image, so called Visual Question Answering (VQA). There are two main modules in algorithm. Given a natural language first module takes then outputs basic questions question. The second question, these We formulate generation problem LASSO optimization problem, also propose criterion how to exploit help Our method is evaluated on challenging VQA dataset yields state-of-the-art accuracy,...

Visual Question Answering (VQA) models should have both high robustness and accuracy. Unfortunately, most of the current VQA research only focuses on accuracy because there is a lack proper methods to measure models. There are two main modules in our algorithm. Given natural language question about an image, first module takes as input then outputs ranked basic questions, with similarity scores, given question. The second question, image these questions text-based answer image. We claim that...

Automatically generating medical reports from retinal images is a difficult task in which an algorithm must generate semantically coherent descriptions for given image. Existing methods mainly rely on the input image to descriptions. However, many abstract concepts or cannot be generated based information only. In this work, we integrate additional help solve task; observe that early diagnosis process, ophthalmologists have usually written down small set of keywords denoting important...

Evaluating the quality of automatically generated image descriptions is challenging, requiring metrics that capture various aspects such as grammaticality, coverage, correctness, and truthfulness. While human evaluation offers valuable insights, its cost time-consuming nature pose limitations. Existing automated like BLEU, ROUGE, METEOR, CIDEr aim to bridge this gap but often show weak correlations with judgment. We address challenge by introducing a novel framework rooted in modern large...

Large language models (LLMs) often produce unsupported or unverifiable information, known as "hallucinations." To mitigate this, retrieval-augmented LLMs incorporate citations, grounding the content in verifiable sources. Despite such developments, manually assessing how well a citation supports associated statement remains major challenge. Previous studies use faithfulness metrics to estimate support automatically but are limited binary classification, overlooking fine-grained practical...

Recently, video summarization has been proposed as a method to help exploration. However, traditional models only generate fixed summary which is usually independent of user-specific needs and hence limits the effectiveness Multi-modal one approaches utilized address this issue. input text-based query input. Hence, effective modeling interaction between essential multi-modal summarization. In work, new causality-based named Causal Video Summarizer (CVS) effectively capture interactive...