Multi-modal large language models (MLLMs) have shown impressive capabilities as a general-purpose interface for various visual and linguistic tasks. However, building a unified MLLM for multi-task learning in the medical field remains a thorny challenge. To mitigate the tug-of-war problem of multi-modal multi-task optimization in MLLMs, recent advances primarily focus on improving the LLM components, while neglecting the connector that bridges the gap between modalities. In this paper, we introduce Uni-Med, a novel medical generalist foundation model which consists of a universal visual feature extraction module, a connector mixture-of-experts (CMoE) module, and an LLM. Benefiting from the proposed CMoE that leverages a well-designed router with a mixture of projection experts at the connector, Uni-Med achieves efficient solution to the tug-of-war problem and can perform six different medical tasks including question answering, visual question answering, report generation, referring expression comprehension, referring expression generation and image classification. To the best of our knowledge, Uni-Med is the first effort to tackle multi-task interference at the connector in MLLMs. Extensive ablation experiments validate the effectiveness of introducing CMoE under any configuration, with up to an average 8% performance gains. We further provide interpretation analysis of the tug-of-war problem from the perspective of gradient optimization and parameter statistics. Compared to previous state-of-the-art medical MLLMs, Uni-Med achieves competitive or superior evaluation metrics on diverse tasks. Code and resources are available at https://github.com/tsinghua-msiip/Uni-Med.
Referring Expression Comprehension (REC), which aims to ground a local visual region via natural language, is a task that heavily relies on multimodal alignment. Most existing methods utilize powerful pre-trained models to transfer visual/linguistic knowledge by full fine-tuning. However, full fine-tuning the entire backbone not only breaks the rich prior knowledge embedded in the pre-training, but also incurs significant computational costs. Motivated by the recent emergence of Parameter-Efficient Transfer Learning (PETL) methods, we aim to solve the REC task in an effective and efficient manner. Directly applying these PETL methods to the REC task is inappropriate, as they lack the specific-domain abilities for precise local visual perception and visual-language alignment. Therefore, we propose a novel framework of Multimodal Prior-guided Parameter Efficient Tuning, namely MaPPER. Specifically, MaPPER comprises Dynamic Prior Adapters guided by an aligned prior, and Local Convolution Adapters to extract precise local semantics for better visual perception. Moreover, the Prior-Guided Text module is proposed to further utilize the prior for facilitating the cross-modal alignment. Experimental results on three widely-used benchmarks demonstrate that MaPPER achieves the best accuracy compared to the full fine-tuning and other PETL methods with only 1.41% tunable backbone parameters. Our code is available at https://github.com/liuting20/MaPPER.
Referring Expression Comprehension (REC) is a crucial cross-modal task that objectively evaluates the capabilities of language understanding, image comprehension, and language-to-image grounding. Consequently, it serves as an ideal testing ground for Multi-modal Large Language Models (MLLMs). In pursuit of this goal, we have established a new REC dataset characterized by two key features: Firstly, it is designed with controllable varying levels of difficulty, necessitating multi-level fine-grained reasoning across object categories, attributes, and multi-hop relationships. Secondly, it includes negative text and images created through fine-grained editing and generation based on existing data, thereby testing the model's ability to correctly reject scenarios where the target object is not visible in the image--an essential aspect often overlooked in existing datasets and approaches. Utilizing this high-quality dataset, we conducted comprehensive evaluations of both state-of-the-art specialist models and MLLMs. Our findings indicate that there remains a significant gap in achieving satisfactory grounding performance. We anticipate that our dataset will inspire new approaches to enhance visual reasoning and develop more advanced cross-modal interaction strategies, ultimately unlocking the full potential of MLLMs. Our code and the datasets are available at https://github.com/liujunzhuo/FineCops-Ref.
Vision Language Models (VLMs) have demonstrated remarkable capabilities in various open-vocabulary tasks, yet their zero-shot performance lags behind task-specific finetuned models, particularly in complex tasks like Referring Expression Comprehension (REC). Fine-tuning usually requires 'white-box' access to the model's architecture and weights, which is not always feasible due to proprietary or privacy concerns. In this work, we propose LLM-wrapper, a method for 'black-box' adaptation of VLMs for the REC task using Large Language Models (LLMs). LLM-wrapper capitalizes on the reasoning abilities of LLMs, improved with a light fine-tuning, to select the most relevant bounding box matching the referring expression, from candidates generated by a zero-shot black-box VLM. Our approach offers several advantages: it enables the adaptation of closed-source models without needing access to their internal workings, it is versatile as it works with any VLM, it transfers to new VLMs, and it allows for the adaptation of an ensemble of VLMs. We evaluate LLM-wrapper on multiple datasets using different VLMs and LLMs, demonstrating significant performance improvements and highlighting the versatility of our method. While LLM-wrapper is not meant to directly compete with standard white-box fine-tuning, it offers a practical and effective alternative for black-box VLM adaptation. The code will be open-sourced.
One common belief is that with complex models and pre-training on large-scale datasets, transformer-based methods for referring expression comprehension (REC) perform much better than existing graph-based methods. We observe that since most graph-based methods adopt an off-the-shelf detector to locate candidate objects (i.e., regions detected by the object detector), they face two challenges that result in subpar performance: (1) the presence of significant noise caused by numerous irrelevant objects during reasoning, and (2) inaccurate localization outcomes attributed to the provided detector. To address these issues, we introduce a plug-and-adapt module guided by sub-expressions, called dynamic gate constraint (DGC), which can adaptively disable irrelevant proposals and their connections in graphs during reasoning. We further introduce an expression-guided regression strategy (EGR) to refine location prediction. Extensive experimental results on the RefCOCO, RefCOCO+, RefCOCOg, Flickr30K, RefClef, and Ref-reasoning datasets demonstrate the effectiveness of the DGC module and the EGR strategy in consistently boosting the performances of various graph-based REC methods. Without any pretaining, the proposed graph-based method achieves better performance than the state-of-the-art (SOTA) transformer-based methods.
Object detection is a fundamental challenge in computer vision, centered on recognizing objects within images, with diverse applications in areas like image analysis, robotics, and autonomous vehicles. Although existing methods have achieved great success, they are often constrained by a fixed vocabulary of objects. To overcome this limitation, approaches like MDETR have redefined object detection by incorporating region-level vision-language pre-training, enabling open-vocabulary object detectors. However, these methods are computationally heavy due to the simultaneous training of large models for both vision and language representations. To address this, we introduce a lightweight framework that significantly reduces the number of parameters while preserving, or even improving, performance. Our solution is applied to MDETR, resulting in the development of Lightweight MDETR (LightMDETR), an optimized version of MDETR designed to enhance computational efficiency without sacrificing accuracy. The core of our approach involves freezing the MDETR backbone and training only the Universal Projection module (UP), which bridges vision and language representations. A learnable modality token parameter allows the UP to seamlessly switch between modalities. Evaluations on tasks like phrase grounding, referring expression comprehension, and segmentation show that LightMDETR not only reduces computational costs but also outperforms several state-of-the-art methods in terms of accuracy.
With the advent of multi-modal large language models (MLLMs), datasets used for visual question answering (VQA) and referring expression comprehension have seen a resurgence. However, the most popular datasets used to evaluate MLLMs are some of the earliest ones created, and they have many known problems, including extreme bias, spurious correlations, and an inability to permit fine-grained analysis. In this paper, we pioneer evaluating recent MLLMs (LLaVA 1.5, LLaVA-NeXT, BLIP2, InstructBLIP, GPT-4V, and GPT-4o) on datasets designed to address weaknesses in earlier ones. We assess three VQA datasets: 1) TDIUC, which permits fine-grained analysis on 12 question types; 2) TallyQA, which has simple and complex counting questions; and 3) DVQA, which requires optical character recognition for chart understanding. We also study VQDv1, a dataset that requires identifying all image regions that satisfy a given query. Our experiments reveal the weaknesses of many MLLMs that have not previously been reported. Our code is integrated into the widely used LAVIS framework for MLLM evaluation, enabling the rapid assessment of future MLLMs. Project webpage: https://kevinlujian.github.io/MLLM_Evaluations/
Vision-language foundation models such as CLIP have achieved tremendous results in global vision-language alignment, but still show some limitations in creating representations for specific image regions. % To address this problem, we propose MaskInversion, a method that leverages the feature representations of pre-trained foundation models, such as CLIP, to generate a context-aware embedding for a query image region specified by a mask at test time. MaskInversion starts with initializing an embedding token and compares its explainability map, derived from the foundation model, to the query mask. The embedding token is then subsequently refined to approximate the query region by minimizing the discrepancy between its explainability map and the query mask. During this process, only the embedding vector is updated, while the underlying foundation model is kept frozen allowing to use MaskInversion with any pre-trained model. As deriving the explainability map involves computing its gradient, which can be expensive, we propose a gradient decomposition strategy that simplifies this computation. The learned region representation can be used for a broad range of tasks, including open-vocabulary class retrieval, referring expression comprehension, as well as for localized captioning and image generation. We evaluate the proposed method on all those tasks on several datasets such as PascalVOC, MSCOCO, RefCOCO, and OpenImagesV7 and show its capabilities compared to other SOTA approaches.
Visual grounding tasks aim to localize image regions based on natural language references. In this work, we explore whether generative VLMs predominantly trained on image-text data could be leveraged to scale up the text annotation of visual grounding data. We find that grounding knowledge already exists in generative VLM and can be elicited by proper prompting. We thus prompt a VLM to generate object-level descriptions by feeding it object regions from existing object detection datasets. We further propose attribute modeling to explicitly capture the important object attributes, and spatial relation modeling to capture inter-object relationship, both of which are common linguistic pattern in referring expression. Our constructed dataset (500K images, 1M objects, 16M referring expressions) is one of the largest grounding datasets to date, and the first grounding dataset with purely model-generated queries and human-annotated objects. To verify the quality of this data, we conduct zero-shot transfer experiments to the popular RefCOCO benchmarks for both referring expression comprehension (REC) and segmentation (RES) tasks. On both tasks, our model significantly outperform the state-of-the-art approaches without using human annotated visual grounding data. Our results demonstrate the promise of generative VLM to scale up visual grounding in the real world. Code and models will be released.
3D referring expression comprehension (3DREC) and segmentation (3DRES) have overlapping objectives, indicating their potential for collaboration. However, existing collaborative approaches predominantly depend on the results of one task to make predictions for the other, limiting effective collaboration. We argue that employing separate branches for 3DREC and 3DRES tasks enhances the model's capacity to learn specific information for each task, enabling them to acquire complementary knowledge. Thus, we propose the MCLN framework, which includes independent branches for 3DREC and 3DRES tasks. This enables dedicated exploration of each task and effective coordination between the branches. Furthermore, to facilitate mutual reinforcement between these branches, we introduce a Relative Superpoint Aggregation (RSA) module and an Adaptive Soft Alignment (ASA) module. These modules significantly contribute to the precise alignment of prediction results from the two branches, directing the module to allocate increased attention to key positions. Comprehensive experimental evaluation demonstrates that our proposed method achieves state-of-the-art performance on both the 3DREC and 3DRES tasks, with an increase of 2.05% in [email omitted] for 3DREC and 3.96% in mIoU for 3DRES.
This report presents a solution for the zero-shot referring expression comprehension task. Visual-language multimodal base models (such as CLIP, SAM) have gained significant attention in recent years as a cornerstone of mainstream research. One of the key applications of multimodal base models lies in their ability to generalize to zero-shot downstream tasks. Unlike traditional referring expression comprehension, zero-shot referring expression comprehension aims to apply pre-trained visual-language models directly to the task without specific training. Recent studies have enhanced the zero-shot performance of multimodal base models in referring expression comprehension tasks by introducing visual prompts. To address the zero-shot referring expression comprehension challenge, we introduced a combination of visual prompts and considered the influence of textual prompts, employing joint prediction tailored to the data characteristics. Ultimately, our approach achieved accuracy rates of 84.825 on the A leaderboard and 71.460 on the B leaderboard, securing the first position.
Referring expression comprehension (REC) is a vision-language task to locate a target object in an image based on a language expression. Fully fine-tuning general-purpose pre-trained vision-language foundation models for REC yields impressive performance but becomes increasingly costly. Parameter-efficient transfer learning (PETL) methods have shown strong performance with fewer tunable parameters. However, directly applying PETL to REC faces two challenges: (1) insufficient multi-modal interaction between pre-trained vision-language foundation models, and (2) high GPU memory usage due to gradients passing through the heavy vision-language foundation models. To this end, we present M$^2$IST: Multi-Modal Interactive Side-Tuning with M$^3$ISAs: Mixture of Multi-Modal Interactive Side-Adapters. During fine-tuning, we keep the pre-trained uni-modal encoders fixed, updating M$^3$ISAs on side networks to progressively connect them, enabling more comprehensive vision-language alignment and efficient tuning for REC. Empirical results reveal that M$^2$IST achieves an optimal balance between performance and efficiency compared to most full fine-tuning and other PETL methods. With M$^2$IST, standard transformer-based REC methods present competitive or even superior performance compared to full fine-tuning, while utilizing only 2.11\% of the tunable parameters, 39.61\% of the GPU memory, and 63.46\% of the fine-tuning time required for full fine-tuning.
Grounding DINO and the Segment Anything Model (SAM) have achieved impressive performance in zero-shot object detection and image segmentation, respectively. Together, they have a great potential to revolutionize applications in zero-shot semantic segmentation or data annotation. Yet, in specialized domains like medical image segmentation, objects of interest (e.g., organs, tissues, and tumors) may not fall in existing class names. To address this problem, the referring expression comprehension (REC) ability of Grounding DINO is leveraged to detect arbitrary targets by their language descriptions. However, recent studies have highlighted severe limitation of the REC framework in this application setting owing to its tendency to make false positive predictions when the target is absent in the given image. And, while this bottleneck is central to the prospect of open-set semantic segmentation, it is still largely unknown how much improvement can be achieved by studying the prediction errors. To this end, we perform empirical studies on six publicly available datasets across different domains and reveal that these errors consistently follow a predictable pattern and can, thus, be mitigated by a simple strategy. Specifically, we show that false positive detections with appreciable confidence scores generally occupy large image areas and can usually be filtered by their relative sizes. More importantly, we expect these observations to inspire future research in improving REC-based detection and automated segmentation. Meanwhile, we evaluate the performance of SAM on multiple datasets from various specialized domains and report significant improvements in segmentation performance and annotation time savings over manual approaches.
Referring Expression Comprehension (REC) aims to localize the target objects specified by free-form natural language descriptions in images. While state-of-the-art methods achieve impressive performance, they perform a dense perception of images, which incorporates redundant visual regions unrelated to linguistic queries, leading to additional computational overhead. This inspires us to explore a question: can we eliminate linguistic-irrelevant redundant visual regions to improve the efficiency of the model? Existing relevant methods primarily focus on fundamental visual tasks, with limited exploration in vision-language fields. To address this, we propose a coarse-to-fine iterative perception framework, called ScanFormer. It can iteratively exploit the image scale pyramid to extract linguistic-relevant visual patches from top to bottom. In each iteration, irrelevant patches are discarded by our designed informativeness prediction. Furthermore, we propose a patch selection strategy for discarded patches to accelerate inference. Experiments on widely used datasets, namely RefCOCO, RefCOCO+, RefCOCOg, and ReferItGame, verify the effectiveness of our method, which can strike a balance between accuracy and efficiency.
Referring expression comprehension (REC) involves localizing a target instance based on a textual description. Recent advancements in REC have been driven by large multimodal models (LMMs) like CogVLM, which achieved 92.44% accuracy on RefCOCO. However, this study questions whether existing benchmarks such as RefCOCO, RefCOCO+, and RefCOCOg, capture LMMs' comprehensive capabilities. We begin with a manual examination of these benchmarks, revealing high labeling error rates: 14% in RefCOCO, 24% in RefCOCO+, and 5% in RefCOCOg, which undermines the authenticity of evaluations. We address this by excluding problematic instances and reevaluating several LMMs capable of handling the REC task, showing significant accuracy improvements, thus highlighting the impact of benchmark noise. In response, we introduce Ref-L4, a comprehensive REC benchmark, specifically designed to evaluate modern REC models. Ref-L4 is distinguished by four key features: 1) a substantial sample size with 45,341 annotations; 2) a diverse range of object categories with 365 distinct types and varying instance scales from 30 to 3,767; 3) lengthy referring expressions averaging 24.2 words; and 4) an extensive vocabulary comprising 22,813 unique words. We evaluate a total of 24 large models on Ref-L4 and provide valuable insights. The cleaned versions of RefCOCO, RefCOCO+, and RefCOCOg, as well as our Ref-L4 benchmark and evaluation code, are available at https://github.com/JierunChen/Ref-L4.
Embodied perception is essential for intelligent vehicles and robots in interactive environmental understanding. However, these advancements primarily focus on vision, with limited attention given to using 3D modeling sensors, restricting a comprehensive understanding of objects in response to prompts containing qualitative and quantitative queries. Recently, as a promising automotive sensor with affordable cost, 4D millimeter-wave radars provide denser point clouds than conventional radars and perceive both semantic and physical characteristics of objects, thereby enhancing the reliability of perception systems. To foster the development of natural language-driven context understanding in radar scenes for 3D visual grounding, we construct the first dataset, Talk2Radar, which bridges these two modalities for 3D Referring Expression Comprehension (REC). Talk2Radar contains 8,682 referring prompt samples with 20,558 referred objects. Moreover, we propose a novel model, T-RadarNet, for 3D REC on point clouds, achieving State-Of-The-Art (SOTA) performance on the Talk2Radar dataset compared to counterparts. Deformable-FPN and Gated Graph Fusion are meticulously designed for efficient point cloud feature modeling and cross-modal fusion between radar and text features, respectively. Comprehensive experiments provide deep insights into radar-based 3D REC. We release our project at https://github.com/GuanRunwei/Talk2Radar.
Multi-modal Large Language Models (MLLMs) have recently achieved enhanced performance across various vision-language tasks including visual grounding capabilities. However, the adversarial robustness of visual grounding remains unexplored in MLLMs. To fill this gap, we use referring expression comprehension (REC) as an example task in visual grounding and propose three adversarial attack paradigms as follows. Firstly, untargeted adversarial attacks induce MLLMs to generate incorrect bounding boxes for each object. Besides, exclusive targeted adversarial attacks cause all generated outputs to the same target bounding box. In addition, permuted targeted adversarial attacks aim to permute all bounding boxes among different objects within a single image. Extensive experiments demonstrate that the proposed methods can successfully attack visual grounding capabilities of MLLMs. Our methods not only provide a new perspective for designing novel attacks but also serve as a strong baseline for improving the adversarial robustness for visual grounding of MLLMs.
Visual affordance learning is a key component for robots to understand how to interact with objects. Conventional approaches in this field rely on pre-defined objects and actions, falling short of capturing diverse interactions in realworld scenarios. The key idea of our approach is employing textual instruction, targeting various affordances for a wide range of objects. This approach covers both hand-object and tool-object interactions. We introduce text-driven affordance learning, aiming to learn contact points and manipulation trajectories from an egocentric view following textual instruction. In our task, contact points are represented as heatmaps, and the manipulation trajectory as sequences of coordinates that incorporate both linear and rotational movements for various manipulations. However, when we gather data for this task, manual annotations of these diverse interactions are costly. To this end, we propose a pseudo dataset creation pipeline and build a large pseudo-training dataset: TextAFF80K, consisting of over 80K instances of the contact points, trajectories, images, and text tuples. We extend existing referring expression comprehension models for our task, and experimental results show that our approach robustly handles multiple affordances, serving as a new standard for affordance learning in real-world scenarios.
In this work, we introduce Mini-Gemini, a simple and effective framework enhancing multi-modality Vision Language Models (VLMs). Despite the advancements in VLMs facilitating basic visual dialog and reasoning, a performance gap persists compared to advanced models like GPT-4 and Gemini. We try to narrow the gap by mining the potential of VLMs for better performance and any-to-any workflow from three aspects, i.e., high-resolution visual tokens, high-quality data, and VLM-guided generation. To enhance visual tokens, we propose to utilize an additional visual encoder for high-resolution refinement without increasing the visual token count. We further construct a high-quality dataset that promotes precise image comprehension and reasoning-based generation, expanding the operational scope of current VLMs. In general, Mini-Gemini further mines the potential of VLMs and empowers current frameworks with image understanding, reasoning, and generation simultaneously. Mini-Gemini supports a series of dense and MoE Large Language Models (LLMs) from 2B to 34B. It is demonstrated to achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models. Code and models are available at https://github.com/dvlab-research/MiniGemini.
Visual Programming has recently emerged as an alternative to end-to-end black-box visual reasoning models. This type of method leverages Large Language Models (LLMs) to generate the source code for an executable computer program that solves a given problem. This strategy has the advantage of offering an interpretable reasoning path and does not require finetuning a model with task-specific data. We propose PropTest, a general strategy that improves visual programming by further using an LLM to generate code that tests for visual properties in an initial round of proposed solutions. Our method generates tests for data-type consistency, output syntax, and semantic properties. PropTest achieves comparable results to state-of-the-art methods while using publicly available LLMs. This is demonstrated across different benchmarks on visual question answering and referring expression comprehension. Particularly, PropTest improves ViperGPT by obtaining 46.1\% accuracy (+6.0\%) on GQA using Llama3-8B and 59.5\% (+8.1\%) on RefCOCO+ using CodeLlama-34B.