Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning

Generating detailed captions comprehending text-rich visual content in images has received growing attention for Large Vision-Language Models (LVLMs). However, few studies have developed benchmarks specifically tailored for detailed captions to measure their accuracy and comprehensiveness. In this paper, we introduce a detailed caption benchmark, termed as CompreCap, to evaluate the visual context from a directed scene graph view. Concretely, we first manually segment the image into semantically meaningful regions (i.e., semantic segmentation mask) according to common-object vocabulary, while also distinguishing attributes of objects within all those regions. Then directional relation labels of these objects are annotated to compose a directed scene graph that can well encode rich compositional information of the image. Based on our directed scene graph, we develop a pipeline to assess the generated detailed captions from LVLMs on multiple levels, including the object-level coverage, the accuracy of attribute descriptions, the score of key relationships, etc. Experimental results on the CompreCap dataset confirm that our evaluation method aligns closely with human evaluation scores across LVLMs.

Learning Visual Generative Priors without Text

Although text-to-image (T2I) models have recently thrived as visual generative priors, their reliance on high-quality text-image pairs makes scaling up expensive. We argue that grasping the cross-modality alignment is not a necessity for a sound visual generative prior, whose focus should be on texture modeling. Such a philosophy inspires us to study image-to-image (I2I) generation, where models can learn from in-the-wild images in a self-supervised manner. We first develop a pure vision-based training framework, Lumos, and confirm the feasibility and the scalability of learning I2I models. We then find that, as an upstream task of T2I, our I2I model serves as a more foundational visual prior and achieves on-par or better performance than existing T2I models using only 1/10 text-image pairs for fine-tuning. We further demonstrate the superiority of I2I priors over T2I priors on some text-irrelevant visual generative tasks, like image-to-3D and image-to-video.

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.

LoTLIP: Improving Language-Image Pre-training for Long Text Understanding

Understanding long text is of great demands in practice but beyond the reach of most language-image pre-training (LIP) models. In this work, we empirically confirm that the key reason causing such an issue is that the training images are usually paired with short captions, leaving certain tokens easily overshadowed by salient tokens. Towards this problem, our initial attempt is to relabel the data with long captions, however, directly learning with which may lead to performance degradation in understanding short text (e.g., in the image classification task). Then, after incorporating corner tokens to aggregate diverse textual information, we manage to help the model catch up to its original level of short text understanding yet greatly enhance its capability of long text understanding. We further look into whether the model can continuously benefit from longer captions and notice a clear trade-off between the performance and the efficiency. Finally, we validate the effectiveness of our approach using a self-constructed large-scale dataset, which consists of 100M long caption oriented text-image pairs. It is noteworthy that, on the task of long-text image retrieval, we beat the competitor using long captions with 11.1% improvement (i.e., from 72.62% to 83.72%). We will release the code, the model, and the new dataset to facilitate the reproducibility and further research. The project page is available at https://wuw2019.github.io/lotlip.

ControlMTR: Control-Guided Motion Transformer with Scene-Compliant Intention Points for Feasible Motion Prediction

The ability to accurately predict feasible multimodal future trajectories of surrounding traffic participants is crucial for behavior planning in autonomous vehicles. The Motion Transformer (MTR), a state-of-the-art motion prediction method, alleviated mode collapse and instability during training and enhanced overall prediction performance by replacing conventional dense future endpoints with a small set of fixed prior motion intention points. However, the fixed prior intention points make the MTR multi-modal prediction distribution over-scattered and infeasible in many scenarios. In this paper, we propose the ControlMTR framework to tackle the aforementioned issues by generating scene-compliant intention points and additionally predicting driving control commands, which are then converted into trajectories by a simple kinematic model with soft constraints. These control-generated trajectories will guide the directly predicted trajectories by an auxiliary loss function. Together with our proposed scene-compliant intention points, they can effectively restrict the prediction distribution within the road boundaries and suppress infeasible off-road predictions while enhancing prediction performance. Remarkably, without resorting to additional model ensemble techniques, our method surpasses the baseline MTR model across all performance metrics, achieving notable improvements of 5.22% in SoftmAP and a 4.15% reduction in MissRate. Our approach notably results in a 41.85% reduction in the cross-boundary rate of the MTR, effectively ensuring that the prediction distribution is confined within the drivable area.

CoReS: Orchestrating the Dance of Reasoning and Segmentation

The reasoning segmentation task, which demands a nuanced comprehension of intricate queries to accurately pinpoint object regions, is attracting increasing attention. However, Multi-modal Large Language Models (MLLM) often find it difficult to accurately localize the objects described in complex reasoning contexts. We believe that the act of reasoning segmentation should mirror the cognitive stages of human visual search, where each step is a progressive refinement of thought toward the final object. Thus we introduce the Chains of Reasoning and Segmenting (CoReS) and find this top-down visual hierarchy indeed enhances the visual search process. Specifically, we propose a dual-chain structure that generates multi-modal, chain-like outputs to aid the segmentation process. Furthermore, to steer the MLLM's outputs into this intended hierarchy, we incorporate in-context inputs as guidance. Extensive experiments demonstrate the superior performance of our CoReS, which surpasses the state-of-the-art method by 7.1\% on the ReasonSeg dataset. The code will be released at https://github.com/baoxiaoyi/CoReS.

DreamLIP: Language-Image Pre-training with Long Captions

Language-image pre-training largely relies on how precisely and thoroughly a text describes its paired image. In practice, however, the contents of an image can be so rich that well describing them requires lengthy captions (e.g., with 10 sentences), which are usually missing in existing datasets. Consequently, there are currently no clear evidences on whether and how language-image pre-training could benefit from long captions. To figure this out, we first re-caption 30M images with detailed descriptions using a pre-trained Multi-modality Large Language Model (MLLM), and then study the usage of the resulting captions under a contrastive learning framework. We observe that, each sentence within a long caption is very likely to describe the image partially (e.g., an object). Motivated by this, we propose to dynamically sample sub-captions from the text label to construct multiple positive pairs, and introduce a grouping loss to match the embeddings of each sub-caption with its corresponding local image patches in a self-supervised manner. Experimental results on a wide rage of downstream tasks demonstrate the consistent superiority of our method, termed DreamLIP, over previous alternatives, highlighting its fine-grained representational capacity. It is noteworthy that, on the tasks of image-text retrieval and semantic segmentation, our model trained with 30M image-text pairs achieves on par or even better performance than CLIP trained with 400M pairs. Project page is available at https://zyf0619sjtu.github.io/dream-lip.

Understanding the Multi-modal Prompts of the Pre-trained Vision-Language Model

Prompt learning has emerged as an efficient alternative for fine-tuning foundational models, such as CLIP, for various downstream tasks. However, there is no work that provides a comprehensive explanation for the working mechanism of the multi-modal prompts. In this paper, we conduct a direct analysis of the multi-modal prompts by asking the following questions: $(i)$ How do the learned multi-modal prompts improve the recognition performance? $(ii)$ What do the multi-modal prompts learn? To answer these questions, we begin by isolating the component of the formula where the prompt influences the calculation of self-attention at each layer in two distinct ways, \ie, $(1)$ introducing prompt embeddings makes the $[cls]$ token focus on foreground objects. $(2)$ the prompts learn a bias term during the update of token embeddings, allowing the model to adapt to the target domain. Subsequently, we conduct extensive visualization and statistical experiments on the eleven diverse downstream recognition datasets. From the experiments, we reveal that the learned prompts improve the performance mainly through the second way, which acts as the dataset bias to improve the recognition performance of the pre-trained model on the corresponding dataset. Based on this finding, we propose the bias tuning way and demonstrate that directly incorporating the learnable bias outperforms the learnable prompts in the same parameter settings. In datasets with limited category information, \ie, EuroSAT, bias tuning surpasses prompt tuning by a large margin. With a deeper understanding of the multi-modal prompt, we hope our work can inspire new and solid research in this direction.

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.

Detecting the open-world objects with the help of the Brain

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 detect unseen/unknown objects and incrementally learn them. The natural instinct of humans to identify unknown objects in their environments mainly depends on their brains' knowledge base. It is difficult for a model to do this only by learning from the annotation of several tiny datasets. The large pre-trained grounded language-image models - VL (\ie GLIP) have rich knowledge about the open world but are limited to the text prompt. We propose leveraging the VL as the ``Brain'' of the open-world detector by simply generating unknown labels. Leveraging it is non-trivial because the unknown labels impair the model's learning of known objects. In this paper, we alleviate these problems by proposing the down-weight loss function and decoupled detection structure. Moreover, our detector leverages the ``Brain'' to learn novel objects beyond VL through our pseudo-labeling scheme.