Previous works on object detection have achieved high accuracy in closed-set scenarios, but their performance in open-world scenarios is not satisfactory. One of the challenging open-world problems is corner case detection in autonomous driving. Existing detectors struggle with these cases, relying heavily on visual appearance and exhibiting poor generalization ability. In this paper, we propose a solution by reducing the discrepancy between known and unknown classes and introduce a multimodal-enhanced objectness notion learner. Leveraging both vision-centric and image-text modalities, our semi-supervised learning framework imparts objectness knowledge to the student model, enabling class-aware detection. Our approach, Multimodal-Enhanced Objectness Learner (MENOL) for Corner Case Detection, significantly improves recall for novel classes with lower training costs. By achieving a 76.6% mAR-corner and 79.8% mAR-agnostic on the CODA-val dataset with just 5100 labeled training images, MENOL outperforms the baseline ORE by 71.3% and 60.6%, respectively. The code will be available at https://github.com/tryhiseyyysum/MENOL.
Image restoration is a fundamental problem that involves recovering a high-quality clean image from its degraded observation. All-In-One image restoration models can effectively restore images from various types and levels of degradation using degradation-specific information as prompts to guide the restoration model. In this work, we present the first approach that uses human-written instructions to guide the image restoration model. Given natural language prompts, our model can recover high-quality images from their degraded counterparts, considering multiple degradation types. Our method, InstructIR, achieves state-of-the-art results on several restoration tasks including image denoising, deraining, deblurring, dehazing, and (low-light) image enhancement. InstructIR improves +1dB over previous all-in-one restoration methods. Moreover, our dataset and results represent a novel benchmark for new research on text-guided image restoration and enhancement. Our code, datasets and models are available at: https://github.com/mv-lab/InstructIR
The risk of hardware Trojans being inserted at various stages of chip production has increased in a zero-trust fabless era. To counter this, various machine learning solutions have been developed for the detection of hardware Trojans. While most of the focus has been on either a statistical or deep learning approach, the limited number of Trojan-infected benchmarks affects the detection accuracy and restricts the possibility of detecting zero-day Trojans. To close the gap, we first employ generative adversarial networks to amplify our data in two alternative representation modalities, a graph and a tabular, ensuring that the dataset is distributed in a representative manner. Further, we propose a multimodal deep learning approach to detect hardware Trojans and evaluate the results from both early fusion and late fusion strategies. We also estimate the uncertainty quantification metrics of each prediction for risk-aware decision-making. The outcomes not only confirms the efficacy of our proposed hardware Trojan detection method but also opens a new door for future studies employing multimodality and uncertainty quantification to address other hardware security challenges.
This work introduces an enhanced approach to generating scene graphs by incorporating both a relationship hierarchy and commonsense knowledge. Specifically, we begin by proposing a hierarchical relation head that exploits an informative hierarchical structure. It jointly predicts the relation super-category between object pairs in an image, along with detailed relations under each super-category. Following this, we implement a robust commonsense validation pipeline that harnesses foundation models to critique the results from the scene graph prediction system, removing nonsensical predicates even with a small language-only model. Extensive experiments on Visual Genome and OpenImage V6 datasets demonstrate that the proposed modules can be seamlessly integrated as plug-and-play enhancements to existing scene graph generation algorithms. The results show significant improvements with an extensive set of reasonable predictions beyond dataset annotations. Codes are available at https://github.com/bowen-upenn/scene_graph_commonsense.
Technological advances in medical data collection, such as high-throughput genomic sequencing and digital high-resolution histopathology, have contributed to the rising requirement for multimodal biomedical modelling, specifically for image, tabular and graph data. Most multimodal deep learning approaches use modality-specific architectures that are often trained separately and cannot capture the crucial cross-modal information that motivates the integration of different data sources. This paper presents the Hybrid Early-fusion Attention Learning Network (HEALNet): a flexible multimodal fusion architecture, which a) preserves modality-specific structural information, b) captures the cross-modal interactions and structural information in a shared latent space, c) can effectively handle missing modalities during training and inference, and d) enables intuitive model inspection by learning on the raw data input instead of opaque embeddings. We conduct multimodal survival analysis on Whole Slide Images and Multi-omic data on four cancer datasets from The Cancer Genome Atlas (TCGA). HEALNet achieves state-of-the-art performance compared to other end-to-end trained fusion models, substantially improving over unimodal and multimodal baselines whilst being robust in scenarios with missing modalities.
The versatility of multimodal deep learning holds tremendous promise for advancing scientific research and practical applications. As this field continues to evolve, the collective power of cross-modal analysis promises to drive transformative innovations, opening new frontiers in chemical understanding and drug discovery. Hence, we introduce Asymmetric Contrastive Multimodal Learning (ACML), a specifically designed approach to enhance molecular understanding and accelerate advancements in drug discovery. ACML harnesses the power of effective asymmetric contrastive learning to seamlessly transfer information from various chemical modalities to molecular graph representations. By combining pre-trained chemical unimodal encoders and a shallow-designed graph encoder with 5 layers, ACML facilitates the assimilation of coordinated chemical semantics from different modalities, leading to comprehensive representation learning with efficient training. We demonstrate the effectiveness of this framework through large-scale cross-modality retrieval and isomer discrimination tasks. Additionally, ACML enhances interpretability by revealing chemical semantics in graph presentations and bolsters the expressive power of graph neural networks, as evidenced by improved performance in molecular property prediction tasks from MoleculeNet and Therapeutics Data Commons (TDC). Ultimately, ACML exemplifies its potential to revolutionize molecular representational learning, offering deeper insights into the chemical semantics of diverse modalities and paving the way for groundbreaking advancements in chemical research and drug discovery.
This paper explores post-disaster analytics using multimodal deep learning models trained with curriculum learning method. Studying post-disaster analytics is important as it plays a crucial role in mitigating the impact of disasters by providing timely and accurate insights into the extent of damage and the allocation of resources. We propose a curriculum learning strategy to enhance the performance of multimodal deep learning models. Curriculum learning emulates the progressive learning sequence in human education by training deep learning models on increasingly complex data. Our primary objective is to develop a curriculum-trained multimodal deep learning model, with a particular focus on visual question answering (VQA) capable of jointly processing image and text data, in conjunction with semantic segmentation for disaster analytics using the FloodNet\footnote{https://github.com/BinaLab/FloodNet-Challenge-EARTHVISION2021} dataset. To achieve this, U-Net model is used for semantic segmentation and image encoding. A custom built text classifier is used for visual question answering. Existing curriculum learning methods rely on manually defined difficulty functions. We introduce a novel curriculum learning approach termed Dynamic Task and Weight Prioritization (DATWEP), which leverages a gradient-based method to automatically decide task difficulty during curriculum learning training, thereby eliminating the need for explicit difficulty computation. The integration of DATWEP into our multimodal model shows improvement on VQA performance. Source code is available at https://github.com/fualsan/DATWEP.
The video-language (VL) pretraining has achieved remarkable improvement in multiple downstream tasks. However, the current VL pretraining framework is hard to extend to multiple modalities (N modalities, N>=3) beyond vision and language. We thus propose LanguageBind, taking the language as the bind across different modalities because the language modality is well-explored and contains rich semantics. Specifically, we freeze the language encoder acquired by VL pretraining, then train encoders for other modalities with contrastive learning. As a result, all modalities are mapped to a shared feature space, implementing multi-modal semantic alignment. While LanguageBind ensures that we can extend VL modalities to N modalities, we also need a high-quality dataset with alignment data pairs centered on language. We thus propose VIDAL-10M with Video, Infrared, Depth, Audio and their corresponding Language, naming as VIDAL-10M. In our VIDAL-10M, all videos are from short video platforms with complete semantics rather than truncated segments from long videos, and all the video, depth, infrared, and audio modalities are aligned to their textual descriptions. LanguageBind has achieved superior performance on a wide range of 15 benchmarks covering video, audio, depth, and infrared. Moreover, multiple experiments have provided evidence for the effectiveness of LanguageBind in achieving indirect alignment and complementarity among diverse modalities. Code address: https://github.com/PKU-YuanGroup/LanguageBind
In recent years, deep learning has shown promise in predicting hypertension (HTN) from fundus images. However, most prior research has primarily focused on analyzing a single type of data, which may not capture the full complexity of HTN risk. To address this limitation, this study introduces a multimodal deep learning (MMDL) system, dubbed HyMNet, which combines fundus images and cardiometabolic risk factors, specifically age and gender, to improve hypertension detection capabilities. Our MMDL system uses RETFound, a foundation model pre-trained on 1.6 million retinal images, for the fundus path and a fully connected neural network for the age and gender path. The two paths are jointly trained by concatenating the feature vectors from each path that are then fed into a fusion network. The system was trained on 5,016 retinal images from 1,243 individuals collected from the Saudi Ministry of National Guard Health Affairs. The results show that the multimodal model that integrates fundus images along with age and gender outperforms the unimodal system trained solely on fundus photographs, with an F1 score of 0.771 [0.747, 0.796], and 0.745 [0.719, 0.772] for hypertension detection, respectively. Additionally, we studied the effect underlying diabetes mellitus has on the model's predictive ability, concluding that diabetes is used as a confounding variable for distinguishing hypertensive cases. Our code and model weights are publicly available at https://github.com/MohammedSB/HyMNet.
Recommender systems (RSs) offer personalized navigation experiences on online platforms, but recommendation remains a challenging task, particularly in specific scenarios and domains. Multimodality can help tap into richer information sources and construct more refined user/item profiles for recommendations. However, existing literature lacks a shared and universal schema for modeling and solving the recommendation problem through the lens of multimodality. This work aims to formalize a general multimodal schema for multimedia recommendation. It provides a comprehensive literature review of multimodal approaches for multimedia recommendation from the last eight years, outlines the theoretical foundations of a multimodal pipeline, and demonstrates its rationale by applying it to selected state-of-the-art approaches. The work also conducts a benchmarking analysis of recent algorithms for multimedia recommendation within Elliot, a rigorous framework for evaluating recommender systems. The main aim is to provide guidelines for designing and implementing the next generation of multimodal approaches in multimedia recommendation.
Multimodal deep learning, especially vision-language models, have gained significant traction in recent years, greatly improving performance on many downstream tasks, including content moderation and violence detection. However, standard multimodal approaches often assume consistent modalities between training and inference, limiting applications in many real-world use cases, as some modalities may not be available during inference. While existing research mitigates this problem through reconstructing the missing modalities, they unavoidably increase unnecessary computational cost, which could be just as critical, especially for large, deployed infrastructures in industry. To this end, we propose a novel guidance network that promotes knowledge sharing during training, taking advantage of the multimodal representations to train better single-modality models to be used for inference. Real-world experiments in violence detection shows that our proposed framework trains single-modality models that significantly outperform traditionally trained counterparts, while avoiding increases in computational cost for inference.
Multimodal deep learning foundation models can learn the relationship between images and text. In the context of medical imaging, mapping images to language concepts reflects the clinical task of diagnostic image interpretation, however current general-purpose foundation models do not perform well in this context because their training corpus have limited medical text and images. To address this challenge and account for the range of cardiac physiology, we leverage 1,032,975 cardiac ultrasound videos and corresponding expert interpretations to develop EchoCLIP, a multimodal foundation model for echocardiography. EchoCLIP displays strong zero-shot (not explicitly trained) performance in cardiac function assessment (external validation left ventricular ejection fraction mean absolute error (MAE) of 7.1%) and identification of implanted intracardiac devices (areas under the curve (AUC) between 0.84 and 0.98 for pacemakers and artificial heart valves). We also developed a long-context variant (EchoCLIP-R) with a custom echocardiography report text tokenizer which can accurately identify unique patients across multiple videos (AUC of 0.86), identify clinical changes such as orthotopic heart transplants (AUC of 0.79) or cardiac surgery (AUC 0.77), and enable robust image-to-text search (mean cross-modal retrieval rank in the top 1% of candidate text reports). These emergent capabilities can be used for preliminary assessment and summarization of echocardiographic findings.
Multi-modal foundation models combining vision and language models such as Flamingo or GPT-4 have recently gained enormous interest. Alignment of foundation models is used to prevent models from providing toxic or harmful output. While malicious users have successfully tried to jailbreak foundation models, an equally important question is if honest users could be harmed by malicious third-party content. In this paper we show that imperceivable attacks on images in order to change the caption output of a multi-modal foundation model can be used by malicious content providers to harm honest users e.g. by guiding them to malicious websites or broadcast fake information. This indicates that countermeasures to adversarial attacks should be used by any deployed multi-modal foundation model.
Learning multimodal representations involves integrating information from multiple heterogeneous sources of data. In order to accelerate progress towards understudied modalities and tasks while ensuring real-world robustness, we release MultiZoo, a public toolkit consisting of standardized implementations of > 20 core multimodal algorithms and MultiBench, a large-scale benchmark spanning 15 datasets, 10 modalities, 20 prediction tasks, and 6 research areas. Together, these provide an automated end-to-end machine learning pipeline that simplifies and standardizes data loading, experimental setup, and model evaluation. To enable holistic evaluation, we offer a comprehensive methodology to assess (1) generalization, (2) time and space complexity, and (3) modality robustness. MultiBench paves the way towards a better understanding of the capabilities and limitations of multimodal models, while ensuring ease of use, accessibility, and reproducibility. Our toolkits are publicly available, will be regularly updated, and welcome inputs from the community.
The robustness of multimodal deep learning models to realistic changes in the input text is critical for their applicability to important tasks such as text-to-image retrieval and cross-modal entailment. To measure robustness, several existing approaches edit the text data, but do so without leveraging the cross-modal information present in multimodal data. Information from the visual modality, such as color, size, and shape, provide additional attributes that users can include in their inputs. Thus, we propose cross-modal attribute insertions as a realistic perturbation strategy for vision-and-language data that inserts visual attributes of the objects in the image into the corresponding text (e.g., "girl on a chair" to "little girl on a wooden chair"). Our proposed approach for cross-modal attribute insertions is modular, controllable, and task-agnostic. We find that augmenting input text using cross-modal insertions causes state-of-the-art approaches for text-to-image retrieval and cross-modal entailment to perform poorly, resulting in relative drops of 15% in MRR and 20% in $F_1$ score, respectively. Crowd-sourced annotations demonstrate that cross-modal insertions lead to higher quality augmentations for multimodal data than augmentations using text-only data, and are equivalent in quality to original examples. We release the code to encourage robustness evaluations of deep vision-and-language models: https://github.com/claws-lab/multimodal-robustness-xmai.
Training multimodal networks requires a vast amount of data due to their larger parameter space compared to unimodal networks. Active learning is a widely used technique for reducing data annotation costs by selecting only those samples that could contribute to improving model performance. However, current active learning strategies are mostly designed for unimodal tasks, and when applied to multimodal data, they often result in biased sample selection from the dominant modality. This unfairness hinders balanced multimodal learning, which is crucial for achieving optimal performance. To address this issue, we propose three guidelines for designing a more balanced multimodal active learning strategy. Following these guidelines, a novel approach is proposed to achieve more fair data selection by modulating the gradient embedding with the dominance degree among modalities. Our studies demonstrate that the proposed method achieves more balanced multimodal learning by avoiding greedy sample selection from the dominant modality. Our approach outperforms existing active learning strategies on a variety of multimodal classification tasks. Overall, our work highlights the importance of balancing sample selection in multimodal active learning and provides a practical solution for achieving more balanced active learning for multimodal classification.
We present ImageBind, an approach to learn a joint embedding across six different modalities - images, text, audio, depth, thermal, and IMU data. We show that all combinations of paired data are not necessary to train such a joint embedding, and only image-paired data is sufficient to bind the modalities together. ImageBind can leverage recent large scale vision-language models, and extends their zero-shot capabilities to new modalities just by using their natural pairing with images. It enables novel emergent applications 'out-of-the-box' including cross-modal retrieval, composing modalities with arithmetic, cross-modal detection and generation. The emergent capabilities improve with the strength of the image encoder and we set a new state-of-the-art on emergent zero-shot recognition tasks across modalities, outperforming specialist supervised models. Finally, we show strong few-shot recognition results outperforming prior work, and that ImageBind serves as a new way to evaluate vision models for visual and non-visual tasks.
This work aims to create a multimodal AI system that chats with humans and shares relevant photos. While earlier works were limited to dialogues about specific objects or scenes within images, recent works have incorporated images into open-domain dialogues. However, their response generators are unimodal, accepting text input but no image input, thus prone to generating responses contradictory to the images shared in the dialogue. Therefore, this work proposes a complete chatbot system using two multimodal deep learning models: an image retriever that understands texts and a response generator that understands images. The image retriever, implemented by ViT and BERT, selects the most relevant image given the dialogue history and a database of images. The response generator, implemented by ViT and GPT-2/DialoGPT, generates an appropriate response given the dialogue history and the most recently retrieved image. The two models are trained and evaluated on PhotoChat, an open-domain dialogue dataset in which a photo is shared in each session. In automatic evaluation, the proposed image retriever outperforms existing baselines VSE++ and SCAN with Recall@1/5/10 of 0.1/0.3/0.4 and MRR of 0.2 when ranking 1,000 images. The proposed response generator also surpasses the baseline Divter with PPL of 16.9, BLEU-1/2 of 0.13/0.03, and Distinct-1/2 of 0.97/0.86, showing a significant improvement in PPL by -42.8 and BLEU-1/2 by +0.07/0.02. In human evaluation with a Likert scale of 1-5, the complete multimodal chatbot system receives higher image-groundedness of 4.3 and engagingness of 4.3, along with competitive fluency of 4.1, coherence of 3.9, and humanness of 3.1, when compared to other chatbot variants. The source code is available at: https://github.com/minniie/multimodal_chat.git.
The rise in loosely-structured data available through text, images, and other modalities has called for new ways of querying them. Multimedia Information Retrieval has filled this gap and has witnessed exciting progress in recent years. Tasks such as search and retrieval of extensive multimedia archives have undergone massive performance improvements, driven to a large extent by recent developments in multimodal deep learning. However, methods in this field remain limited in the kinds of queries they support and, in particular, their inability to answer database-like queries. For this reason, inspired by recent work on neural databases, we propose a new framework, which we name Multimodal Neural Databases (MMNDBs). MMNDBs can answer complex database-like queries that involve reasoning over different input modalities, such as text and images, at scale. In this paper, we present the first architecture able to fulfill this set of requirements and test it with several baselines, showing the limitations of currently available models. The results show the potential of these new techniques to process unstructured data coming from different modalities, paving the way for future research in the area. Code to replicate the experiments will be released at https://github.com/GiovanniTRA/MultimodalNeuralDatabases
We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge. With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters. Besides language commands, our approach can be simply extended to multi-modal instructions for learning image-conditioned LLaMA model, which achieves superior reasoning performance on ScienceQA and COCO Caption benchmarks. Furthermore, we also evaluate the zero-initialized attention mechanism for fine-tuning other pre-trained models (ViT, RoBERTa) on traditional vision and language tasks, demonstrating the superior generalization capacity of our approach. Code is released at https://github.com/OpenGVLab/LLaMA-Adapter.