Over the past few years, the advancement of Multimodal Large Language Models (MLLMs) has captured the wide interest of researchers, leading to numerous innovations to enhance MLLMs' comprehension. In this paper, we present AdaptVision, a multimodal large language model specifically designed to dynamically process input images at varying resolutions. We hypothesize that the requisite number of visual tokens for the model is contingent upon both the resolution and content of the input image. Generally, natural images with a lower information density can be effectively interpreted by the model using fewer visual tokens at reduced resolutions. In contrast, images containing textual content, such as documents with rich text, necessitate a higher number of visual tokens for accurate text interpretation due to their higher information density. Building on this insight, we devise a dynamic image partitioning module that adjusts the number of visual tokens according to the size and aspect ratio of images. This method mitigates distortion effects that arise from resizing images to a uniform resolution and dynamically optimizing the visual tokens input to the LLMs. Our model is capable of processing images with resolutions up to $1008\times 1008$. Extensive experiments across various datasets demonstrate that our method achieves impressive performance in handling vision-language tasks in both natural and text-related scenes. The source code and dataset are now publicly available at \url{https://github.com/harrytea/AdaptVision}.
Recent evaluations of Large Multimodal Models (LMMs) have explored their capabilities in various domains, with only few benchmarks specifically focusing on urban environments. Moreover, existing urban benchmarks have been limited to evaluating LMMs with basic region-level urban tasks under singular views, leading to incomplete evaluations of LMMs' abilities in urban environments. To address these issues, we present UrBench, a comprehensive benchmark designed for evaluating LMMs in complex multi-view urban scenarios. UrBench contains 11.6K meticulously curated questions at both region-level and role-level that cover 4 task dimensions: Geo-Localization, Scene Reasoning, Scene Understanding, and Object Understanding, totaling 14 task types. In constructing UrBench, we utilize data from existing datasets and additionally collect data from 11 cities, creating new annotations using a cross-view detection-matching method. With these images and annotations, we then integrate LMM-based, rule-based, and human-based methods to construct large-scale high-quality questions. Our evaluations on 21 LMMs show that current LMMs struggle in the urban environments in several aspects. Even the best performing GPT-4o lags behind humans in most tasks, ranging from simple tasks such as counting to complex tasks such as orientation, localization and object attribute recognition, with an average performance gap of 17.4%. Our benchmark also reveals that LMMs exhibit inconsistent behaviors with different urban views, especially with respect to understanding cross-view relations. UrBench datasets and benchmark results will be publicly available at https://opendatalab.github.io/UrBench/.
Robust semantic segmentation of intraoperative image data holds promise for enabling automatic surgical scene understanding and autonomous robotic surgery. While model development and validation are primarily conducted on idealistic scenes, geometric domain shifts, such as occlusions of the situs, are common in real-world open surgeries. To close this gap, we (1) present the first analysis of state-of-the-art (SOA) semantic segmentation models when faced with geometric out-of-distribution (OOD) data, and (2) propose an augmentation technique called "Organ Transplantation", to enhance generalizability. Our comprehensive validation on six different OOD datasets, comprising 600 RGB and hyperspectral imaging (HSI) cubes from 33 pigs, each annotated with 19 classes, reveals a large performance drop in SOA organ segmentation models on geometric OOD data. This performance decline is observed not only in conventional RGB data (with a dice similarity coefficient (DSC) drop of 46 %) but also in HSI data (with a DSC drop of 45 %), despite the richer spectral information content. The performance decline increases with the spatial granularity of the input data. Our augmentation technique improves SOA model performance by up to 67 % for RGB data and 90 % for HSI data, achieving performance at the level of in-distribution performance on real OOD test data. Given the simplicity and effectiveness of our augmentation method, it is a valuable tool for addressing geometric domain shifts in surgical scene segmentation, regardless of the underlying model. Our code and pre-trained models are publicly available at https://github.com/IMSY-DKFZ/htc.
Reliable embodied perception from an egocentric perspective is challenging yet essential for autonomous navigation technology of intelligent mobile agents. With the growing demand of social robotics, near-field scene understanding becomes an important research topic in the areas of egocentric perceptual tasks related to navigation in both crowded and unstructured environments. Due to the complexity of environmental conditions and difficulty of surrounding obstacles owing to truncation and occlusion, the perception capability under this circumstance is still inferior. To further enhance the intelligence of mobile robots, in this paper, we setup an egocentric multi-sensor data collection platform based on 3 main types of sensors (Camera, LiDAR and Fisheye), which supports flexible sensor configurations to enable dynamic sight of view from ego-perspective, capturing either near or farther areas. Meanwhile, a large-scale multimodal dataset is constructed, named RoboSense, to facilitate egocentric robot perception. Specifically, RoboSense contains more than 133K synchronized data with 1.4M 3D bounding box and IDs annotated in the full $360^{\circ}$ view, forming 216K trajectories across 7.6K temporal sequences. It has $270\times$ and $18\times$ as many annotations of surrounding obstacles within near ranges as the previous datasets collected for autonomous driving scenarios such as KITTI and nuScenes. Moreover, we define a novel matching criterion for near-field 3D perception and prediction metrics. Based on RoboSense, we formulate 6 popular tasks to facilitate the future research development, where the detailed analysis as well as benchmarks are also provided accordingly. Data desensitization measures have been conducted for privacy protection.
Abundant, well-annotated multimodal data in remote sensing are pivotal for aligning complex visual remote sensing (RS) scenes with human language, enabling the development of specialized vision language models across diverse RS interpretation tasks. However, annotating RS images with rich linguistic semantics at scale demands expertise in RS and substantial human labor, making it costly and often impractical. In this study, we propose a workflow that leverages large language models (LLMs) to generate multimodal datasets with semantically rich captions at scale from plain OpenStreetMap (OSM) data for images sourced from the Google Earth Engine (GEE) platform. This approach facilitates the generation of paired remote sensing data and can be readily scaled up using openly available data. Within this framework, we present RSTeller, a multimodal dataset comprising over 1.3 million RS images, each accompanied by two descriptive captions. Extensive experiments demonstrate that RSTeller enhances the performance of multiple existing vision language models for RS scene understanding through continual pre-training. Our methodology significantly reduces the manual effort and expertise needed for annotating remote sensing imagery while democratizing access to high-quality annotated data. This advancement fosters progress in visual language modeling and encourages broader participation in remote sensing research and applications. The RSTeller dataset is available at https://github.com/SlytherinGe/RSTeller.
Multi-task dense scene understanding, which trains a model for multiple dense prediction tasks, has a wide range of application scenarios. Capturing long-range dependency and enhancing cross-task interactions are crucial to multi-task dense prediction. In this paper, we propose MTMamba++, a novel architecture for multi-task scene understanding featuring with a Mamba-based decoder. It contains two types of core blocks: self-task Mamba (STM) block and cross-task Mamba (CTM) block. STM handles long-range dependency by leveraging state-space models, while CTM explicitly models task interactions to facilitate information exchange across tasks. We design two types of CTM block, namely F-CTM and S-CTM, to enhance cross-task interaction from feature and semantic perspectives, respectively. Experiments on NYUDv2, PASCAL-Context, and Cityscapes datasets demonstrate the superior performance of MTMamba++ over CNN-based and Transformer-based methods. The code is available at https://github.com/EnVision-Research/MTMamba.
Text recognition in the wild is an important technique for digital maps and urban scene understanding, in which the natural resembling properties between glyphs is one of the major reasons that lead to wrong recognition results. To address this challenge, we introduce two extremely fine-grained visual recognition benchmark datasets that contain very challenging resembling glyphs (characters/letters) in the wild to be distinguished. Moreover, we propose a simple yet effective two-stage contrastive learning approach to the extremely fine-grained recognition task of resembling glyphs discrimination. In the first stage, we utilize supervised contrastive learning to leverage label information to warm-up the backbone network. In the second stage, we introduce CCFG-Net, a network architecture that integrates classification and contrastive learning in both Euclidean and Angular spaces, in which contrastive learning is applied in both supervised learning and pairwise discrimination manners to enhance the model's feature representation capability. Overall, our proposed approach effectively exploits the complementary strengths of contrastive learning and classification, leading to improved recognition performance on the resembling glyphs. Comparative evaluations with state-of-the-art fine-grained classification approaches under both Convolutional Neural Network (CNN) and Transformer backbones demonstrate the superiority of our proposed method.
This paper explores scene affinity (AIScene), namely intra-scene consistency and inter-scene correlation, for semi-supervised LiDAR semantic segmentation in driving scenes. Adopting teacher-student training, AIScene employs a teacher network to generate pseudo-labeled scenes from unlabeled data, which then supervise the student network's learning. Unlike most methods that include all points in pseudo-labeled scenes for forward propagation but only pseudo-labeled points for backpropagation, AIScene removes points without pseudo-labels, ensuring consistency in both forward and backward propagation within the scene. This simple point erasure strategy effectively prevents unsupervised, semantically ambiguous points (excluded in backpropagation) from affecting the learning of pseudo-labeled points. Moreover, AIScene incorporates patch-based data augmentation, mixing multiple scenes at both scene and instance levels. Compared to existing augmentation techniques that typically perform scene-level mixing between two scenes, our method enhances the semantic diversity of labeled (or pseudo-labeled) scenes, thereby improving the semi-supervised performance of segmentation models. Experiments show that AIScene outperforms previous methods on two popular benchmarks across four settings, achieving notable improvements of 1.9% and 2.1% in the most challenging 1% labeled data.
Open-vocabulary 3D scene understanding (OV-3D) aims to localize and classify novel objects beyond the closed object classes. However, existing approaches and benchmarks primarily focus on the open vocabulary problem within the context of object classes, which is insufficient to provide a holistic evaluation to what extent a model understands the 3D scene. In this paper, we introduce a more challenging task called Generalized Open-Vocabulary 3D Scene Understanding (GOV-3D) to explore the open vocabulary problem beyond object classes. It encompasses an open and diverse set of generalized knowledge, expressed as linguistic queries of fine-grained and object-specific attributes. To this end, we contribute a new benchmark named OpenScan, which consists of 3D object attributes across eight representative linguistic aspects, including affordance, property, material, and more. We further evaluate state-of-the-art OV-3D methods on our OpenScan benchmark, and discover that these methods struggle to comprehend the abstract vocabularies of the GOV-3D task, a challenge that cannot be addressed by simply scaling up object classes during training. We highlight the limitations of existing methodologies and explore a promising direction to overcome the identified shortcomings. Data and code are available at https://github.com/YoujunZhao/OpenScan
Existing top-performance autonomous driving systems typically rely on the multi-modal fusion strategy for reliable scene understanding. This design is however fundamentally restricted due to overlooking the modality-specific strengths and finally hampering the model performance. To address this limitation, in this work, we introduce a novel modality interaction strategy that allows individual per-modality representations to be learned and maintained throughout, enabling their unique characteristics to be exploited during the whole perception pipeline. To demonstrate the effectiveness of the proposed strategy, we design DeepInteraction++, a multi-modal interaction framework characterized by a multi-modal representational interaction encoder and a multi-modal predictive interaction decoder. Specifically, the encoder is implemented as a dual-stream Transformer with specialized attention operation for information exchange and integration between separate modality-specific representations. Our multi-modal representational learning incorporates both object-centric, precise sampling-based feature alignment and global dense information spreading, essential for the more challenging planning task. The decoder is designed to iteratively refine the predictions by alternately aggregating information from separate representations in a unified modality-agnostic manner, realizing multi-modal predictive interaction. Extensive experiments demonstrate the superior performance of the proposed framework on both 3D object detection and end-to-end autonomous driving tasks. Our code is available at https://github.com/fudan-zvg/DeepInteraction.
This paper introduces a novel method for open-vocabulary 3D scene querying in autonomous driving by combining Language Embedded 3D Gaussians with Large Language Models (LLMs). We propose utilizing LLMs to generate both contextually canonical phrases and helping positive words for enhanced segmentation and scene interpretation. Our method leverages GPT-3.5 Turbo as an expert model to create a high-quality text dataset, which we then use to fine-tune smaller, more efficient LLMs for on-device deployment. Our comprehensive evaluation on the WayveScenes101 dataset demonstrates that LLM-guided segmentation significantly outperforms traditional approaches based on predefined canonical phrases. Notably, our fine-tuned smaller models achieve performance comparable to larger expert models while maintaining faster inference times. Through ablation studies, we discover that the effectiveness of helping positive words correlates with model scale, with larger models better equipped to leverage additional semantic information. This work represents a significant advancement towards more efficient, context-aware autonomous driving systems, effectively bridging 3D scene representation with high-level semantic querying while maintaining practical deployment considerations.
Human-object interactions (HOI) detection aims at capturing human-object pairs in images and corresponding actions. It is an important step toward high-level visual reasoning and scene understanding. However, due to the natural bias from the real world, existing methods mostly struggle with rare human-object pairs and lead to sub-optimal results. Recently, with the development of the generative model, a straightforward approach is to construct a more balanced dataset based on a group of supplementary samples. Unfortunately, there is a significant domain gap between the generated data and the original data, and simply merging the generated images into the original dataset cannot significantly boost the performance. To alleviate the above problem, we present a novel model-agnostic framework called \textbf{C}ontext-\textbf{E}nhanced \textbf{F}eature \textbf{A}lignment (CEFA) module, which can effectively align the generated data with the original data at the feature level and bridge the domain gap. Specifically, CEFA consists of a feature alignment module and a context enhancement module. On one hand, considering the crucial role of human-object pairs information in HOI tasks, the feature alignment module aligns the human-object pairs by aggregating instance information. On the other hand, to mitigate the issue of losing important context information caused by the traditional discriminator-style alignment method, we employ a context-enhanced image reconstruction module to improve the model's learning ability of contextual cues. Extensive experiments have shown that our method can serve as a plug-and-play module to improve the detection performance of HOI models on rare categories\footnote{https://github.com/LijunZhang01/CEFA}.
Learning object-centric representations from unsupervised videos is challenging. Unlike most previous approaches that focus on decomposing 2D images, we present a 3D generative model named DynaVol-S for dynamic scenes that enables object-centric learning within a differentiable volume rendering framework. The key idea is to perform object-centric voxelization to capture the 3D nature of the scene, which infers per-object occupancy probabilities at individual spatial locations. These voxel features evolve through a canonical-space deformation function and are optimized in an inverse rendering pipeline with a compositional NeRF. Additionally, our approach integrates 2D semantic features to create 3D semantic grids, representing the scene through multiple disentangled voxel grids. DynaVol-S significantly outperforms existing models in both novel view synthesis and unsupervised decomposition tasks for dynamic scenes. By jointly considering geometric structures and semantic features, it effectively addresses challenging real-world scenarios involving complex object interactions. Furthermore, once trained, the explicitly meaningful voxel features enable additional capabilities that 2D scene decomposition methods cannot achieve, such as novel scene generation through editing geometric shapes or manipulating the motion trajectories of objects.
As machine learning becomes increasingly integral to autonomous decision-making processes involving human interaction, the necessity of comprehending the model's outputs through conversational means increases. Most recently, foundation models are being explored for their potential as post hoc explainers, providing a pathway to elucidate the decision-making mechanisms of predictive models. In this work, we introduce traceable question-answering, leveraging an external knowledge repository to inform the responses of Large Language Models (LLMs) to user queries within a scene understanding task. This knowledge repository comprises contextual details regarding the model's output, containing high-level features, feature importance, and alternative probabilities. We employ subtractive counterfactual reasoning to compute feature importance, a method that entails analysing output variations resulting from decomposing semantic features. Furthermore, to maintain a seamless conversational flow, we integrate four key characteristics - social, causal, selective, and contrastive - drawn from social science research on human explanations into a single-shot prompt, guiding the response generation process. Our evaluation demonstrates that explanations generated by the LLMs encompassed these elements, indicating its potential to bridge the gap between complex model outputs and natural language expressions.
Surgical instrument segmentation is crucial in surgical scene understanding, thereby facilitating surgical safety. Existing algorithms directly detected all instruments of pre-defined categories in the input image, lacking the capability to segment specific instruments according to the surgeon's intention. During different stages of surgery, surgeons exhibit varying preferences and focus toward different surgical instruments. Therefore, an instrument segmentation algorithm that adheres to the surgeon's intention can minimize distractions from irrelevant instruments and assist surgeons to a great extent. The recent Segment Anything Model (SAM) reveals the capability to segment objects following prompts, but the manual annotations for prompts are impractical during the surgery. To address these limitations in operating rooms, we propose an audio-driven surgical instrument segmentation framework, named ASI-Seg, to accurately segment the required surgical instruments by parsing the audio commands of surgeons. Specifically, we propose an intention-oriented multimodal fusion to interpret the segmentation intention from audio commands and retrieve relevant instrument details to facilitate segmentation. Moreover, to guide our ASI-Seg segment of the required surgical instruments, we devise a contrastive learning prompt encoder to effectively distinguish the required instruments from the irrelevant ones. Therefore, our ASI-Seg promotes the workflow in the operating rooms, thereby providing targeted support and reducing the cognitive load on surgeons. Extensive experiments are performed to validate the ASI-Seg framework, which reveals remarkable advantages over classical state-of-the-art and medical SAMs in both semantic segmentation and intention-oriented segmentation. The source code is available at https://github.com/Zonmgin-Zhang/ASI-Seg.
Scene understanding plays an important role in several high-level computer vision applications, such as autonomous vehicles, intelligent video surveillance, or robotics. However, too few solutions have been proposed for indoor/outdoor scene classification to ensure scene context adaptability for computer vision frameworks. We propose the first Lightweight Hybrid Graph Convolutional Neural Network (LH-GCNN)-CNN framework as an add-on to object detection models. The proposed approach uses the output of the CNN object detection model to predict the observed scene type by generating a coherent GCNN representing the semantic and geometric content of the observed scene. This new method, applied to natural scenes, achieves an efficiency of over 90\% for scene classification in a COCO-derived dataset containing a large number of different scenes, while requiring fewer parameters than traditional CNN methods. For the benefit of the scientific community, we will make the source code publicly available: https://github.com/Aymanbegh/Hybrid-GCNN-CNN.
Domain adaptive panoptic segmentation promises to resolve the long tail of corner cases in natural scene understanding. Previous state of the art addresses this problem with cross-task consistency, careful system-level optimization and heuristic improvement of teacher predictions. In contrast, we propose to build upon remarkable capability of mask transformers to estimate their own prediction uncertainty. Our method avoids noise amplification by leveraging fine-grained confidence of panoptic teacher predictions. In particular, we modulate the loss with mask-wide confidence and discourage back-propagation in pixels with uncertain teacher or confident student. Experimental evaluation on standard benchmarks reveals a substantial contribution of the proposed selection techniques. We report 47.4 PQ on Synthia to Cityscapes, which corresponds to an improvement of 6.2 percentage points over the state of the art. The source code is available at https://github.com/helen1c/MC-PanDA.
3D object detection is an indispensable component for scene understanding. However, the annotation of large-scale 3D datasets requires significant human effort. To tackle this problem, many methods adopt weakly supervised 3D object detection that estimates 3D boxes by leveraging 2D boxes and scene/class-specific priors. However, these approaches generally depend on sophisticated manual priors, which is hard to generalize to novel categories and scenes. In this paper, we are motivated to propose a general approach, which can be easily adapted to new scenes and/or classes. A unified framework is developed for learning 3D object detectors from RGB images and associated 2D boxes. In specific, we propose three general components: prior injection module to obtain general object geometric priors from LLM model, 2D space projection constraint to minimize the discrepancy between the boundaries of projected 3D boxes and their corresponding 2D boxes on the image plane, and 3D space geometry constraint to build a Point-to-Box alignment loss to further refine the pose of estimated 3D boxes. Experiments on KITTI and SUN-RGBD datasets demonstrate that our method yields surprisingly high-quality 3D bounding boxes with only 2D annotation. The source code is available at https://github.com/gwenzhang/GGA.
Specular highlight removal plays a pivotal role in multimedia applications, as it enhances the quality and interpretability of images and videos, ultimately improving the performance of downstream tasks such as content-based retrieval, object recognition, and scene understanding. Despite significant advances in deep learning-based methods, current state-of-the-art approaches often rely on additional priors or supervision, limiting their practicality and generalization capability. In this paper, we propose the Dual-Hybrid Attention Network for Specular Highlight Removal (DHAN-SHR), an end-to-end network that introduces novel hybrid attention mechanisms to effectively capture and process information across different scales and domains without relying on additional priors or supervision. DHAN-SHR consists of two key components: the Adaptive Local Hybrid-Domain Dual Attention Transformer (L-HD-DAT) and the Adaptive Global Dual Attention Transformer (G-DAT). The L-HD-DAT captures local inter-channel and inter-pixel dependencies while incorporating spectral domain features, enabling the network to effectively model the complex interactions between specular highlights and the underlying surface properties. The G-DAT models global inter-channel relationships and long-distance pixel dependencies, allowing the network to propagate contextual information across the entire image and generate more coherent and consistent highlight-free results. To evaluate the performance of DHAN-SHR and facilitate future research in this area, we compile a large-scale benchmark dataset comprising a diverse range of images with varying levels of specular highlights. Through extensive experiments, we demonstrate that DHAN-SHR outperforms 18 state-of-the-art methods both quantitatively and qualitatively, setting a new standard for specular highlight removal in multimedia applications.
3D surface reconstruction from multi-view images is essential for scene understanding and interaction. However, complex indoor scenes pose challenges such as ambiguity due to limited observations. Recent implicit surface representations, such as Neural Radiance Fields (NeRFs) and signed distance functions (SDFs), employ various geometric priors to resolve the lack of observed information. Nevertheless, their performance heavily depends on the quality of the pre-trained geometry estimation models. To ease such dependence, we propose regularizing the geometric modeling by explicitly encouraging the mutual information among surface normals of highly correlated scene points. In this way, the geometry learning process is modulated by the second-order correlations from noisy (first-order) geometric priors, thus eliminating the bias due to poor generalization. Additionally, we introduce a simple yet effective scheme that utilizes semantic and geometric features to identify correlated points, enhancing their mutual information accordingly. The proposed technique can serve as a plugin for SDF-based neural surface representations. Our experiments demonstrate the effectiveness of the proposed in improving the surface reconstruction quality of major states of the arts. Our code is available at: \url{https://github.com/Muliphein/InfoNorm}.