LSceneLLM: Enhancing Large 3D Scene Understanding Using Adaptive Visual Preferences

Research on 3D Vision-Language Models (3D-VLMs) is gaining increasing attention, which is crucial for developing embodied AI within 3D scenes, such as visual navigation and embodied question answering. Due to the high density of visual features, especially in large 3D scenes, accurately locating task-relevant visual information is challenging. Existing works attempt to segment all objects and consider their features as scene representations. However, these task-agnostic object features include much redundant information and missing details for the task-relevant area. To tackle these problems, we propose LSceneLLM, an adaptive framework that automatically identifies task-relevant areas by leveraging LLM's visual preference for different tasks, followed by a plug-and-play scene magnifier module to capture fine-grained details in focused areas. Specifically, a dense token selector examines the attention map of LLM to identify visual preferences for the instruction input. It then magnifies fine-grained details of the focusing area. An adaptive self-attention module is leveraged to fuse the coarse-grained and selected fine-grained visual information. To comprehensively evaluate the large scene understanding ability of 3D-VLMs, we further introduce a cross-room understanding benchmark, XR-Scene, which contains a series of large scene understanding tasks including XR-QA, XR-EmbodiedPlanning, and XR-SceneCaption. Experiments show that our method surpasses existing methods on both large scene understanding and existing scene understanding benchmarks. Plunging our scene magnifier module into the existing 3D-VLMs also brings significant improvement.

CoNav: A Benchmark for Human-Centered Collaborative Navigation

Human-robot collaboration, in which the robot intelligently assists the human with the upcoming task, is an appealing objective. To achieve this goal, the agent needs to be equipped with a fundamental collaborative navigation ability, where the agent should reason human intention by observing human activities and then navigate to the human's intended destination in advance of the human. However, this vital ability has not been well studied in previous literature. To fill this gap, we propose a collaborative navigation (CoNav) benchmark. Our CoNav tackles the critical challenge of constructing a 3D navigation environment with realistic and diverse human activities. To achieve this, we design a novel LLM-based humanoid animation generation framework, which is conditioned on both text descriptions and environmental context. The generated humanoid trajectory obeys the environmental context and can be easily integrated into popular simulators. We empirically find that the existing navigation methods struggle in CoNav task since they neglect the perception of human intention. To solve this problem, we propose an intention-aware agent for reasoning both long-term and short-term human intention. The agent predicts navigation action based on the predicted intention and panoramic observation. The emergent agent behavior including observing humans, avoiding human collision, and navigation reveals the efficiency of the proposed datasets and agents.

A Simple Knowledge Distillation Framework for Open-world Object Detection

Open World Object Detection (OWOD) is a novel computer vision task with a considerable challenge, bridging the gap between classic object detection (OD) benchmarks and real-world object detection. In addition to detecting and classifying seen/known objects, OWOD algorithms are expected to localize all potential unseen/unknown objects and incrementally learn them. The large pre-trained vision-language grounding models (VLM,eg, GLIP) have rich knowledge about the open world, but are limited by text prompts and cannot localize indescribable objects. However, there are many detection scenarios which pre-defined language descriptions are unavailable during inference. In this paper, we attempt to specialize the VLM model for OWOD task by distilling its open-world knowledge into a language-agnostic detector. Surprisingly, we observe that the combination of a simple knowledge distillation approach and the automatic pseudo-labeling mechanism in OWOD can achieve better performance for unknown object detection, even with a small amount of data. Unfortunately, knowledge distillation for unknown objects severely affects the learning of detectors with conventional structures for known objects, leading to catastrophic forgetting. To alleviate these problems, we propose the down-weight loss function for knowledge distillation from vision-language to single vision modality. Meanwhile, we decouple the learning of localization and recognition to reduce the impact of category interactions of known and unknown objects on the localization learning process. Comprehensive experiments performed on MS-COCO and PASCAL VOC demonstrate the effectiveness of our methods.

Contrastive Vision-Language Alignment Makes Efficient Instruction Learner

We study the task of extending the large language model (LLM) into a vision-language instruction-following model. This task is crucial but challenging since the LLM is trained on text modality only, making it hard to effectively digest the visual modality. To address this, existing methods typically train a visual adapter to align the representation between a pre-trained vision transformer (ViT) and the LLM by a generative image captioning loss. However, we find that the generative objective can only produce weak alignment for vision and language, making the aligned vision-language model very hungry for the instruction fine-tuning data. In this paper, we propose CG-VLM that applies both Contrastive and Generative alignment objectives to effectively align the representation of ViT and LLM. Different from image level and sentence level alignment in common contrastive learning settings, CG-VLM aligns the image-patch level features and text-token level embeddings, which, however, is very hard to achieve as no explicit grounding patch-token relation provided in standard image captioning datasets. To address this issue, we propose to maximize the averaged similarity between pooled image-patch features and text-token embeddings. Extensive experiments demonstrate that the proposed CG-VLM produces strong vision-language alignment and is an efficient instruction learner. For example, using only 10% instruction tuning data, we reach 95% performance of state-of-the-art method LLaVA [29] on the zero-shot ScienceQA-Image benchmark.

RGM: A Robust Generalist Matching Model

Finding corresponding pixels within a pair of images is a fundamental computer vision task with various applications. Due to the specific requirements of different tasks like optical flow estimation and local feature matching, previous works are primarily categorized into dense matching and sparse feature matching focusing on specialized architectures along with task-specific datasets, which may somewhat hinder the generalization performance of specialized models. In this paper, we propose a deep model for sparse and dense matching, termed RGM (Robust Generalist Matching). In particular, we elaborately design a cascaded GRU module for refinement by exploring the geometric similarity iteratively at multiple scales following an additional uncertainty estimation module for sparsification. To narrow the gap between synthetic training samples and real-world scenarios, we build a new, large-scale dataset with sparse correspondence ground truth by generating optical flow supervision with greater intervals. As such, we are able to mix up various dense and sparse matching datasets, significantly improving the training diversity. The generalization capacity of our proposed RGM is greatly improved by learning the matching and uncertainty estimation in a two-stage manner on the large, mixed data. Superior performance is achieved for zero-shot matching and downstream geometry estimation across multiple datasets, outperforming the previous methods by a large margin.

FGPrompt: Fine-grained Goal Prompting for Image-goal Navigation

Learning to navigate to an image-specified goal is an important but challenging task for autonomous systems. The agent is required to reason the goal location from where a picture is shot. Existing methods try to solve this problem by learning a navigation policy, which captures semantic features of the goal image and observation image independently and lastly fuses them for predicting a sequence of navigation actions. However, these methods suffer from two major limitations. 1) They may miss detailed information in the goal image, and thus fail to reason the goal location. 2) More critically, it is hard to focus on the goal-relevant regions in the observation image, because they attempt to understand observation without goal conditioning. In this paper, we aim to overcome these limitations by designing a Fine-grained Goal Prompting (FGPrompt) method for image-goal navigation. In particular, we leverage fine-grained and high-resolution feature maps in the goal image as prompts to perform conditioned embedding, which preserves detailed information in the goal image and guides the observation encoder to pay attention to goal-relevant regions. Compared with existing methods on the image-goal navigation benchmark, our method brings significant performance improvement on 3 benchmark datasets (i.e., Gibson, MP3D, and HM3D). Especially on Gibson, we surpass the state-of-the-art success rate by 8% with only 1/50 model size. Project page: https://xinyusun.github.io/fgprompt-pages

$A^2$Nav: Action-Aware Zero-Shot Robot Navigation by Exploiting Vision-and-Language Ability of Foundation Models

We study the task of zero-shot vision-and-language navigation (ZS-VLN), a practical yet challenging problem in which an agent learns to navigate following a path described by language instructions without requiring any path-instruction annotation data. Normally, the instructions have complex grammatical structures and often contain various action descriptions (e.g., "proceed beyond", "depart from"). How to correctly understand and execute these action demands is a critical problem, and the absence of annotated data makes it even more challenging. Note that a well-educated human being can easily understand path instructions without the need for any special training. In this paper, we propose an action-aware zero-shot VLN method ($A^2$Nav) by exploiting the vision-and-language ability of foundation models. Specifically, the proposed method consists of an instruction parser and an action-aware navigation policy. The instruction parser utilizes the advanced reasoning ability of large language models (e.g., GPT-3) to decompose complex navigation instructions into a sequence of action-specific object navigation sub-tasks. Each sub-task requires the agent to localize the object and navigate to a specific goal position according to the associated action demand. To accomplish these sub-tasks, an action-aware navigation policy is learned from freely collected action-specific datasets that reveal distinct characteristics of each action demand. We use the learned navigation policy for executing sub-tasks sequentially to follow the navigation instruction. Extensive experiments show $A^2$Nav achieves promising ZS-VLN performance and even surpasses the supervised learning methods on R2R-Habitat and RxR-Habitat datasets.

M$^3$Video: Masked Motion Modeling for Self-Supervised Video Representation Learning

We study self-supervised video representation learning that seeks to learn video features from unlabeled videos, which is widely used for video analysis as labeling videos is labor-intensive. Current methods often mask some video regions and then train a model to reconstruct spatial information in these regions (e.g., original pixels). However, the model is easy to reconstruct this information by considering content in a single frame. As a result, it may neglect to learn the interactions between frames, which are critical for video analysis. In this paper, we present a new self-supervised learning task, called Masked Motion Modeling (M$^3$Video), for learning representation by enforcing the model to predict the motion of moving objects in the masked regions. To generate motion targets for this task, we track the objects using optical flow. The motion targets consist of position transitions and shape changes of the tracked objects, thus the model has to consider multiple frames comprehensively. Besides, to help the model capture fine-grained motion details, we enforce the model to predict trajectory motion targets in high temporal resolution based on a video in low temporal resolution. After pre-training using our M$^3$Video task, the model is able to anticipate fine-grained motion details even taking a sparsely sampled video as input. We conduct extensive experiments on four benchmark datasets. Remarkably, when doing pre-training with 400 epochs, we improve the accuracy from 67.6\% to 69.2\% and from 78.8\% to 79.7\% on Something-Something V2 and Kinetics-400 datasets, respectively.

Descriptellation: Deep Learned Constellation Descriptors for SLAM

Current global localization descriptors in Simultaneous Localization and Mapping (SLAM) often fail under vast viewpoint or appearance changes. Adding topological information of semantic objects into the descriptors ameliorates the problem. However, hand-crafted topological descriptors extract limited information and they are not robust to environmental noise, drastic perspective changes, or object occlusion or misdetections. To solve this problem, we formulate a learning-based approach by constructing constellations from semantically meaningful objects and use Deep Graph Convolution Networks to map the constellation representation to a descriptor. We demonstrate the effectiveness of our Deep Learned Constellation Descriptor (Descriptellation) on the Paris-Rue-Lille and IQmulus datasets. Although Descriptellation is trained on randomly generated simulation datasets, it shows good generalization abilities on real-world datasets. Descriptellation outperforms the PointNet and handcrafted constellation descriptors for global localization, and shows robustness against different types of noise.