비전-언어 모델(VLM)은 시각적 정보와 텍스트 정보를 함께 이해하여 시각적 질의응답(Visual Question Answering)과 같은 태스크에서 의료 영상 분석에 유망한 성능을 보여왔다. 그러나 기존 혈액학 비전-언어 리소스는 여전히 영어 중심에 머물러 있어, 다국어 의료 환경에서의 적용 가능성이 제한된다. 이러한 과제는 남아시아 전반에 걸쳐, 특히 파키스탄에서 두드러지는데, 이 국가에서는 우르두어가 널리 사용됨에도 불구하고 의료 정보 및 디지털 의료 시스템은 대부분 영어에 의존하고 있다. 이 격차를 조사하기 위해 우리는 의료 전문가들을 대상으로 설문 조사를 실시하였으며, 그 결과 임상 문서와 환자 소통 간에 상당한 언어 불일치가 있음이 확인되었고, 이는 다국어 의료 기술의 필요성을 강조한다. 이러한 한계를 해결하기 위해, 우리는 백혈병 및 정상 백혈구 분석을 위한 임상적으로 검증된 영어-우르두어 이중언어 형태학 인식 VQA 벤치마크인 WBCMor VQA를 소개한다. 본 벤치마크는 LeukemiaAttri 및 WBCAtt 데이터셋의 형태학 인식 주석을 활용하여 구성되었으며, 언어적 일관성과 임상적 정확성을 보장하기 위해 도메인 특화 우르두어 혈액학 사전을 지원한다. 최종 벤치마크는 2만 장의 백혈병 및 정상 단일세포 이미지에 대한 VQA 주석으로서 11만 쌍의 이중언어 질의응답 쌍을 포함한다. 또한, 제안된 벤치마크에서 다수의 오픈소스 VLM을 평가하여 기준선(baseline) 성능을 확립한다. 제안된 리소스는 다국어 의료 환경을 위한 접근 가능하고 임상적으로 적합한 AI 시스템 개발을 촉진하는 것을 목표로 한다.
Perceptual organization remains one of the very few established theories on the human visual system. It underpinned many pre-deep seminal works on segmentation and detection, yet research has seen a rapid decline since the preferential shift to learning deep models. Of the limited attempts, most aimed at interpreting complex visual scenes using perceptual organizational rules. This has however been proven to be sub-optimal, since models were unable to effectively capture the visual complexity in real-world imagery. In this paper, we rejuvenate the study of perceptual organization, by advocating two positional changes: (i) we examine purposefully generated synthetic data, instead of complex real imagery, and (ii) we ask machines to synthesize novel perceptually-valid patterns, instead of explaining existing data. Our overall answer lies with the introduction of a novel visual challenge -- the challenge of perceptual question answering (PQA). Upon observing example perceptual question-answer pairs, the goal for PQA is to solve similar questions by generating answers entirely from scratch (see Figure 1). Our first contribution is therefore the first dataset of perceptual question-answer pairs, each generated specifically for a particular Gestalt principle. We then borrow insights from human psychology to design an agent that casts perceptual organization as a self-attention problem, where a proposed grid-to-grid mapping network directly generates answer patterns from scratch. Experiments show our agent to outperform a selection of naive and strong baselines. A human study however indicates that ours uses astronomically more data to learn when compared to an average human, necessitating future research (with or without our dataset).
Computed Tomography (CT) scan, which produces 3D volumetric medical data that can be viewed as hundreds of cross-sectional images (a.k.a. slices), provides detailed anatomical information for diagnosis. For radiologists, creating CT radiology reports is time-consuming and error-prone. A visual question answering (VQA) system that can answer radiologists' questions about some anatomical regions on the CT scan and even automatically generate a radiology report is urgently needed. However, existing VQA systems cannot adequately handle the CT radiology question answering (CTQA) task for: (1) anatomic complexity makes CT images difficult to understand; (2) spatial relationship across hundreds slices is difficult to capture. To address these issues, this paper proposes CT-Agent, a multimodal agentic framework for CTQA. CT-Agent adopts anatomically independent tools to break down the anatomic complexity; furthermore, it efficiently captures the across-slice spatial relationship with a global-local token compression strategy. Experimental results on two 3D chest CT datasets, CT-RATE and RadGenome-ChestCT, verify the superior performance of CT-Agent.
In this paper, we study the problem of visual question answering (VQA) where the image and query are represented by ASP programs that lack domain data. We provide an approach that is orthogonal and complementary to existing knowledge augmentation techniques where we abduce domain relationships of image constructs from past examples. After framing the abduction problem, we provide a baseline approach, and an implementation that significantly improves the accuracy of query answering yet requires few examples.
Visual Question Answering (VQA) is fundamentally compositional in nature, and many questions are simply answered by decomposing them into modular sub-problems. The recent proposed Neural Module Network (NMN) employ this strategy to question answering, whereas heavily rest with off-the-shelf layout parser or additional expert policy regarding the network architecture design instead of learning from the data. These strategies result in the unsatisfactory adaptability to the semantically-complicated variance of the inputs, thereby hindering the representational capacity and generalizability of the model. To tackle this problem, we propose a Semantic-aware modUlar caPsulE Routing framework, termed as SUPER, to better capture the instance-specific vision-semantic characteristics and refine the discriminative representations for prediction. Particularly, five powerful specialized modules as well as dynamic routers are tailored in each layer of the SUPER network, and the compact routing spaces are constructed such that a variety of customizable routes can be sufficiently exploited and the vision-semantic representations can be explicitly calibrated. We comparatively justify the effectiveness and generalization ability of our proposed SUPER scheme over five benchmark datasets, as well as the parametric-efficient advantage. It is worth emphasizing that this work is not to pursue the state-of-the-art results in VQA. Instead, we expect that our model is responsible to provide a novel perspective towards architecture learning and representation calibration for VQA.
Large Vision-Language Models (LVLMs) have shown significant potential in assisting medical diagnosis by leveraging extensive biomedical datasets. However, the advancement of medical image understanding and reasoning critically depends on building high-quality visual instruction data, which is costly and labor-intensive to obtain, particularly in the medical domain. To mitigate this data-starving issue, we introduce Self-Training Large Language and Vision Assistant for Medicine (STLLaVA-Med). The proposed method is designed to train a policy model (an LVLM) capable of auto-generating medical visual instruction data to improve data efficiency, guided through Direct Preference Optimization (DPO). Specifically, a more powerful and larger LVLM (e.g., GPT-4o) is involved as a biomedical expert to oversee the DPO fine-tuning process on the auto-generated data, encouraging the policy model to align efficiently with human preferences. We validate the efficacy and data efficiency of STLLaVA-Med across three major medical Visual Question Answering (VQA) benchmarks, demonstrating competitive zero-shot performance with the utilization of only 9% of the medical data.
We propose a series of recurrent and contextual neural network models for multiple choice visual question answering on the Visual7W dataset. Motivated by divergent trends in model complexities in the literature, we explore the balance between model expressiveness and simplicity by studying incrementally more complex architectures. We start with LSTM-encoding of input questions and answers; build on this with context generation by LSTM-encodings of neural image and question representations and attention over images; and evaluate the diversity and predictive power of our models and the ensemble thereof. All models are evaluated against a simple baseline inspired by the current state-of-the-art, consisting of involving simple concatenation of bag-of-words and CNN representations for the text and images, respectively. Generally, we observe marked variation in image-reasoning performance between our models not obvious from their overall performance, as well as evidence of dataset bias. Our standalone models achieve accuracies up to $64.6\%$, while the ensemble of all models achieves the best accuracy of $66.67\%$, within $0.5\%$ of the current state-of-the-art for Visual7W.
실제 문서에 대한 질문에 답하려면 텍스트와 표가 섞여 있는 긴 입력을 처리해야 합니다. 문맥을 이미지로 표현하는 광학 문맥 압축(optical context compression)은 토큰 비용을 줄일 수 있을 것으로 기대되지만, 표 이해에 미치는 영향은 여전히 불분명합니다. 본 연구에서는 여러 표가 포함된 문서 대상의 질의응답을 위해 픽셀 수준의 표 압축을 연구하며, 2개의 벤치마크와 5가지 시각적 토큰(token) 예산에 걸쳐 5개의 VLM을 평가합니다. 원본 해상도의 이미지로 표를 표현하는 것은 성능과 효율성 모두에서 텍스트와 대등하지만, 이를 축소(downscaling)하면 모델(model)이 가독성의 손실을 더 길고 덜 효과적인 추론 궤적으로 보충하려 하여 예상했던 절감 효과를 상쇄합니다. 그러나 크게 축소된 표라도 질문과의 관련성 여부를 식별하기에는 충분한 신호를 유지합니다. 우리는 이러한 비대칭성을 활용하여 학습(training)이 필요 없는 2단계 방식을 제안합니다. 모델은 먼저 픽셀 압축된 문맥에서 질문에 답하는 데 필요한 표를 식별한 다음, 해당 표를 원본 해상도에서 추론합니다. 긴 문서에서 우리의 방법은 원본 해상도 표를 사용한 단일 단계 QA 대비 전체 토큰의 41%를 절약하고 정확도를 7%p 높였습니다. 또한 정확도 손실 없이 가장 효율적인 단일 단계 압축 구성보다 15% 더 적은 토큰을 사용합니다.
Human-like attention as a supervisory signal to guide neural attention has shown significant promise but is currently limited to uni-modal integration - even for inherently multimodal tasks such as visual question answering (VQA). We present the Multimodal Human-like Attention Network (MULAN) - the first method for multimodal integration of human-like attention on image and text during training of VQA models. MULAN integrates attention predictions from two state-of-the-art text and image saliency models into neural self-attention layers of a recent transformer-based VQA model. Through evaluations on the challenging VQAv2 dataset, we show that MULAN achieves a new state-of-the-art performance of 73.98% accuracy on test-std and 73.72% on test-dev and, at the same time, has approximately 80% fewer trainable parameters than prior work. Overall, our work underlines the potential of integrating multimodal human-like and neural attention for VQA
The traditional image captioning task uses generic reference captions to provide textual information about images. Different user populations, however, will care about different visual aspects of images. In this paper, we propose a new task, Captioning with a Purpose (CapWAP). Our goal is to develop systems that can be tailored to be useful for the information needs of an intended population, rather than merely provide generic information about an image. In this task, we use question-answer (QA) pairs---a natural expression of information need---from users, instead of reference captions, for both training and post-inference evaluation. We show that it is possible to use reinforcement learning to directly optimize for the intended information need, by rewarding outputs that allow a question answering model to provide correct answers to sampled user questions. We convert several visual question answering datasets into CapWAP datasets, and demonstrate that under a variety of scenarios our purposeful captioning system learns to anticipate and fulfill specific information needs better than its generic counterparts, as measured by QA performance on user questions from unseen images, when using the caption alone as context.
Vision-and-language (V\&L) reasoning necessitates perception of visual concepts such as objects and actions, understanding semantics and language grounding, and reasoning about the interplay between the two modalities. One crucial aspect of visual reasoning is spatial understanding, which involves understanding relative locations of objects, i.e.\ implicitly learning the geometry of the scene. In this work, we evaluate the faithfulness of V\&L models to such geometric understanding, by formulating the prediction of pair-wise relative locations of objects as a classification as well as a regression task. Our findings suggest that state-of-the-art transformer-based V\&L models lack sufficient abilities to excel at this task. Motivated by this, we design two objectives as proxies for 3D spatial reasoning (SR) -- object centroid estimation, and relative position estimation, and train V\&L with weak supervision from off-the-shelf depth estimators. This leads to considerable improvements in accuracy for the "GQA" visual question answering challenge (in fully supervised, few-shot, and O.O.D settings) as well as improvements in relative spatial reasoning. Code and data will be released \href{https://github.com/pratyay-banerjee/weak_sup_vqa}{here}.
Visual question answering (VQA) refers to the problem where, given an image and a natural language question about the image, a correct natural language answer has to be generated. A VQA model has to demonstrate both the visual understanding of the image and the semantic understanding of the question, demonstrating reasoning capability. Since the inception of this field, a plethora of VQA datasets and models have been published. In this article, we meticulously analyze the current state of VQA datasets and models, while cleanly dividing them into distinct categories and then summarizing the methodologies and characteristics of each category. We divide VQA datasets into four categories: (1) available datasets that contain a rich collection of authentic images, (2) synthetic datasets that contain only synthetic images produced through artificial means, (3) diagnostic datasets that are specially designed to test model performance in a particular area, e.g., understanding the scene text, and (4) KB (Knowledge-Based) datasets that are designed to measure a model's ability to utilize outside knowledge. Concurrently, we explore six main paradigms of VQA models: fusion, where we discuss different methods of fusing information between visual and textual modalities; attention, the technique of using information from one modality to filter information from another; external knowledge base, where we discuss different models utilizing outside information; composition or reasoning, where we analyze techniques to answer advanced questions that require complex reasoning steps; explanation, which is the process of generating visual and textual descriptions to verify sound reasoning; and graph models, which encode and manipulate relationships through nodes in a graph. We also discuss some miscellaneous topics, such as scene text understanding, counting, and bias reduction.
Incorporating external knowledge to Visual Question Answering (VQA) has become a vital practical need. Existing methods mostly adopt pipeline approaches with different components for knowledge matching and extraction, feature learning, etc.However, such pipeline approaches suffer when some component does not perform well, which leads to error propagation and poor overall performance. Furthermore, the majority of existing approaches ignore the answer bias issue -- many answers may have never appeared during training (i.e., unseen answers) in real-word application. To bridge these gaps, in this paper, we propose a Zero-shot VQA algorithm using knowledge graphs and a mask-based learning mechanism for better incorporating external knowledge, and present new answer-based Zero-shot VQA splits for the F-VQA dataset. Experiments show that our method can achieve state-of-the-art performance in Zero-shot VQA with unseen answers, meanwhile dramatically augment existing end-to-end models on the normal F-VQA task.
For glaciologists, studying ice sheets from the polar regions is critical. With the advancement of deep learning techniques, we can now extract high-level information from the ice sheet data (e.g., estimating the ice layer thickness, predicting the ice accumulation for upcoming years, etc.). However, a vision-based conversational deep learning approach has not been explored yet, where scientists can get information by asking questions about images. In this paper, we have introduced the task of Visual Question Answering (VQA) on remote-sensed ice sheet imagery. To study, we have presented a unique VQA dataset, Polar-VQA, in this study. All the images in this dataset were collected using four types of airborne radars. The main objective of this research is to highlight the importance of VQA in the context of ice sheet research and conduct a baseline study of existing VQA approaches on Polar-VQA dataset.
Vision and language understanding has emerged as a subject undergoing intense study in Artificial Intelligence. Among many tasks in this line of research, visual question answering (VQA) has been one of the most successful ones, where the goal is to learn a model that understands visual content at region-level details and finds their associations with pairs of questions and answers in the natural language form. Despite the rapid progress in the past few years, most existing work in VQA have focused primarily on images. In this paper, we focus on extending VQA to the video domain and contribute to the literature in three important ways. First, we propose three new tasks designed specifically for video VQA, which require spatio-temporal reasoning from videos to answer questions correctly. Next, we introduce a new large-scale dataset for video VQA named TGIF-QA that extends existing VQA work with our new tasks. Finally, we propose a dual-LSTM based approach with both spatial and temporal attention, and show its effectiveness over conventional VQA techniques through empirical evaluations.
Knowledge-based visual question answering (KVQA) has been extensively studied to answer visual questions with external knowledge, e.g., knowledge graphs (KGs). While several attempts have been proposed to leverage large language models (LLMs) as an implicit knowledge source, it remains challenging since LLMs may generate hallucinations. Moreover, multiple knowledge sources, e.g., images, KGs and LLMs, cannot be readily aligned for complex scenarios. To tackle these, we present a novel modality-aware integration with LLMs for KVQA (MAIL). It carefully leverages multimodal knowledge for both image understanding and knowledge reasoning. Specifically, (i) we propose a two-stage prompting strategy with LLMs to densely embody the image into a scene graph with detailed visual features; (ii) We construct a coupled concept graph by linking the mentioned entities with external facts. (iii) A tailored pseudo-siamese graph medium fusion is designed for sufficient multimodal fusion. We utilize the shared mentioned entities in two graphs as mediums to bridge a tight inter-modal exchange, while maximally preserving insightful intra-modal learning by constraining the fusion within mediums. Extensive experiments on two benchmark datasets show the superiority of MAIL with 24x less resources.
Visual Grounding (VG) methods in Visual Question Answering (VQA) attempt to improve VQA performance by strengthening a model's reliance on question-relevant visual information. The presence of such relevant information in the visual input is typically assumed in training and testing. This assumption, however, is inherently flawed when dealing with imperfect image representations common in large-scale VQA, where the information carried by visual features frequently deviates from expected ground-truth contents. As a result, training and testing of VG-methods is performed with largely inaccurate data, which obstructs proper assessment of their potential benefits. In this study, we demonstrate that current evaluation schemes for VG-methods are problematic due to the flawed assumption of availability of relevant visual information. Our experiments show that these methods can be much more effective when evaluation conditions are corrected. Code is provided on GitHub.
When answering questions about an image, it not only needs knowing what -- understanding the fine-grained contents (e.g., objects, relationships) in the image, but also telling why -- reasoning over grounding visual cues to derive the answer for a question. Over the last few years, we have seen significant progress on visual question answering. Though impressive as the accuracy grows, it still lags behind to get knowing whether these models are undertaking grounding visual reasoning or just leveraging spurious correlations in the training data. Recently, a number of works have attempted to answer this question from perspectives such as grounding and robustness. However, most of them are either focusing on the language side or coarsely studying the pixel-level attention maps. In this paper, by leveraging the step-wise object grounding annotations provided in the GQA dataset, we first present a systematical object-centric diagnosis of visual reasoning on grounding and robustness, particularly on the vision side. According to the extensive comparisons across different models, we find that even models with high accuracy are not good at grounding objects precisely, nor robust to visual content perturbations. In contrast, symbolic and modular models have a relatively better grounding and robustness, though at the cost of accuracy. To reconcile these different aspects, we further develop a diagnostic model, namely Graph Reasoning Machine. Our model replaces purely symbolic visual representation with probabilistic scene graph and then applies teacher-forcing training for the visual reasoning module. The designed model improves the performance on all three metrics over the vanilla neural-symbolic model while inheriting the transparency. Further ablation studies suggest that this improvement is mainly due to more accurate image understanding and proper intermediate reasoning supervisions.
Visual question answering (VQA) has been intensively studied as a multimodal task that requires effort in bridging vision and language to infer answers correctly. Recent attempts have developed various attention-based modules for solving VQA tasks. However, the performance of model inference is largely bottlenecked by visual processing for semantics understanding. Most existing detection methods rely on bounding boxes, remaining a serious challenge for VQA models to understand the causal nexus of object semantics in images and correctly infer contextual information. To this end, we propose a finer model framework without bounding boxes in this work, termed Looking Out of Instance Semantics (LOIS) to tackle this important issue. LOIS enables more fine-grained feature descriptions to produce visual facts. Furthermore, to overcome the label ambiguity caused by instance masks, two types of relation attention modules: 1) intra-modality and 2) inter-modality, are devised to infer the correct answers from the different multi-view features. Specifically, we implement a mutual relation attention module to model sophisticated and deeper visual semantic relations between instance objects and background information. In addition, our proposed attention model can further analyze salient image regions by focusing on important word-related questions. Experimental results on four benchmark VQA datasets prove that our proposed method has favorable performance in improving visual reasoning capability.
Visual question answering (VQA) is a fundamental and essential AI task, and VQA-based disaster scenario understanding is a hot research topic. For instance, we can ask questions about a disaster image by the VQA model and the answer can help identify whether anyone or anything is affected by the disaster. However, previous VQA models for disaster damage assessment have some shortcomings, such as limited candidate answer space, monotonous question types, and limited answering capability of existing models. In this paper, we propose a zero-shot VQA model named Zero-shot VQA for Flood Disaster Damage Assessment (ZFDDA). It is a VQA model for damage assessment without pre-training. Also, with flood disaster as the main research object, we build a Freestyle Flood Disaster Image Question Answering dataset (FFD-IQA) to evaluate our VQA model. This new dataset expands the question types to include free-form, multiple-choice, and yes-no questions. At the same time, we expand the size of the previous dataset to contain a total of 2,058 images and 22,422 question-meta ground truth pairs. Most importantly, our model uses well-designed chain of thought (CoT) demonstrations to unlock the potential of the large language model, allowing zero-shot VQA to show better performance in disaster scenarios. The experimental results show that the accuracy in answering complex questions is greatly improved with CoT prompts. Our study provides a research basis for subsequent research of VQA for other disaster scenarios.