The problem of roadside monocular 3D detection requires detecting objects of interested classes in a 2D RGB frame and predicting their 3D information such as locations in bird's-eye-view (BEV). It has broad applications in traffic control, vehicle-vehicle communication, and vehicle-infrastructure cooperative perception. To approach this problem, we present a novel and simple method by prompting the 3D detector using 2D detections. Our method builds on a key insight that, compared with 3D detectors, a 2D detector is much easier to train and performs significantly better w.r.t detections on the 2D image plane. That said, one can exploit 2D detections of a well-trained 2D detector as prompts to a 3D detector, being trained in a way of inflating such 2D detections to 3D towards 3D detection. To construct better prompts using the 2D detector, we explore three techniques: (a) concatenating both 2D and 3D detectors' features, (b) attentively fusing 2D and 3D detectors' features, and (c) encoding predicted 2D boxes x, y, width, height, label and attentively fusing such with the 3D detector's features. Surprisingly, the third performs the best. Moreover, we present a yaw tuning tactic and a class-grouping strategy that merges classes based on their functionality; these techniques improve 3D detection performance further. Comprehensive ablation studies and extensive experiments demonstrate that our method resoundingly outperforms prior works, achieving the state-of-the-art on two large-scale roadside 3D detection benchmarks.
Pre-trained models with large-scale training data, such as CLIP and Stable Diffusion, have demonstrated remarkable performance in various high-level computer vision tasks such as image understanding and generation from language descriptions. Yet, their potential for low-level tasks such as image restoration remains relatively unexplored. In this paper, we explore such models to enhance image restoration. As off-the-shelf features (OSF) from pre-trained models do not directly serve image restoration, we propose to learn an additional lightweight module called Pre-Train-Guided Refinement Module (PTG-RM) to refine restoration results of a target restoration network with OSF. PTG-RM consists of two components, Pre-Train-Guided Spatial-Varying Enhancement (PTG-SVE), and Pre-Train-Guided Channel-Spatial Attention (PTG-CSA). PTG-SVE enables optimal short- and long-range neural operations, while PTG-CSA enhances spatial-channel attention for restoration-related learning. Extensive experiments demonstrate that PTG-RM, with its compact size ($<$1M parameters), effectively enhances restoration performance of various models across different tasks, including low-light enhancement, deraining, deblurring, and denoising.
Vision-language models (VLMs) excel in zero-shot recognition but their performance varies greatly across different visual concepts. For example, although CLIP achieves impressive accuracy on ImageNet (60-80%), its performance drops below 10% for more than ten concepts like night snake, presumably due to their limited presence in the pretraining data. However, measuring the frequency of concepts in VLMs' large-scale datasets is challenging. We address this by using large language models (LLMs) to count the number of pretraining texts that contain synonyms of these concepts. Our analysis confirms that popular datasets, such as LAION, exhibit a long-tailed concept distribution, yielding biased performance in VLMs. We also find that downstream applications of VLMs, including visual chatbots (e.g., GPT-4V) and text-to-image models (e.g., Stable Diffusion), often fail to recognize or generate images of rare concepts identified by our method. To mitigate the imbalanced performance of zero-shot VLMs, we propose REtrieval-Augmented Learning (REAL). First, instead of prompting VLMs using the original class names, REAL uses their most frequent synonyms found in pretraining texts. This simple change already outperforms costly human-engineered and LLM-enriched prompts over nine benchmark datasets. Second, REAL trains a linear classifier on a small yet balanced set of pretraining data retrieved using concept synonyms. REAL surpasses the previous zero-shot SOTA, using 400x less storage and 10,000x less training time!
In our increasingly diverse society, everyday physical interfaces often present barriers, impacting individuals across various contexts. This oversight, from small cabinet knobs to identical wall switches that can pose different contextual challenges, highlights an imperative need for solutions. Leveraging low-cost 3D-printed augmentations such as knob magnifiers and tactile labels seems promising, yet the process of discovering unrecognized barriers remains challenging because disability is context-dependent. We introduce AccessLens, an end-to-end system designed to identify inaccessible interfaces in daily objects, and recommend 3D-printable augmentations for accessibility enhancement. Our approach involves training a detector using the novel AccessDB dataset designed to automatically recognize 21 distinct Inaccessibility Classes (e.g., bar-small and round-rotate) within 6 common object categories (e.g., handle and knob). AccessMeta serves as a robust way to build a comprehensive dictionary linking these accessibility classes to open-source 3D augmentation designs. Experiments demonstrate our detector's performance in detecting inaccessible objects.
Few-shot object detection (FSOD) benchmarks have advanced techniques for detecting new categories with limited annotations. Existing benchmarks repurpose well-established datasets like COCO by partitioning categories into base and novel classes for pre-training and fine-tuning respectively. However, these benchmarks do not reflect how FSOD is deployed in practice. Rather than only pre-training on a small number of base categories, we argue that it is more practical to fine-tune a foundation model (e.g., a vision-language model (VLM) pre-trained on web-scale data) for a target domain. Surprisingly, we find that zero-shot inference from VLMs like GroundingDINO significantly outperforms the state-of-the-art (48.3 vs. 33.1 AP) on COCO. However, such zero-shot models can still be misaligned to target concepts of interest. For example, trailers on the web may be different from trailers in the context of autonomous vehicles. In this work, we propose Foundational FSOD, a new benchmark protocol that evaluates detectors pre-trained on any external datasets and fine-tuned on K-shots per target class. Further, we note that current FSOD benchmarks are actually federated datasets containing exhaustive annotations for each category on a subset of the data. We leverage this insight to propose simple strategies for fine-tuning VLMs with federated losses. We demonstrate the effectiveness of our approach on LVIS and nuImages, improving over prior work by 5.9 AP.
Autonomous vehicles (AVs) must accurately detect objects from both common and rare classes for safe navigation, motivating the problem of Long-Tailed 3D Object Detection (LT3D). Contemporary LiDAR-based 3D detectors perform poorly on rare classes (e.g., CenterPoint only achieves 5.1 AP on stroller) as it is difficult to recognize objects from sparse LiDAR points alone. RGB images provide visual evidence to help resolve such ambiguities, motivating the study of RGB-LiDAR fusion. In this paper, we delve into a simple late-fusion framework that ensembles independently trained RGB and LiDAR detectors. Unlike recent end-to-end methods which require paired multi-modal training data, our late-fusion approach can easily leverage large-scale uni-modal datasets, significantly improving rare class detection.In particular, we examine three critical components in this late-fusion framework from first principles, including whether to train 2D or 3D RGB detectors, whether to match RGB and LiDAR detections in 3D or the projected 2D image plane, and how to fuse matched detections.Extensive experiments reveal that 2D RGB detectors achieve better recognition accuracy than 3D RGB detectors, matching on the 2D image plane mitigates depth estimation errors, and fusing scores probabilistically with calibration leads to state-of-the-art LT3D performance. Our late-fusion approach achieves 51.4 mAP on the established nuScenes LT3D benchmark, improving over prior work by 5.9 mAP.
Contrastive Language-Image Pre-training (CLIP) plays an essential role in extracting valuable content information from images across diverse tasks. It aligns textual and visual modalities to comprehend the entire image, including all the details, even those irrelevant to specific tasks. However, for a finer understanding and controlled editing of images, it becomes crucial to focus on specific regions of interest, which can be indicated as points, masks, or boxes by humans or perception models. To fulfill the requirements, we introduce Alpha-CLIP, an enhanced version of CLIP with an auxiliary alpha channel to suggest attentive regions and fine-tuned with constructed millions of RGBA region-text pairs. Alpha-CLIP not only preserves the visual recognition ability of CLIP but also enables precise control over the emphasis of image contents. It demonstrates effectiveness in various tasks, including but not limited to open-world recognition, multimodal large language models, and conditional 2D / 3D generation. It has a strong potential to serve as a versatile tool for image-related tasks.
Egocentric sensors such as AR/VR devices capture human-object interactions and offer the potential to provide task-assistance by recalling 3D locations of objects of interest in the surrounding environment. This capability requires instance tracking in real-world 3D scenes from egocentric videos (IT3DEgo). We explore this problem by first introducing a new benchmark dataset, consisting of RGB and depth videos, per-frame camera pose, and instance-level annotations in both 2D camera and 3D world coordinates. We present an evaluation protocol which evaluates tracking performance in 3D coordinates with two settings for enrolling instances to track: (1) single-view online enrollment where an instance is specified on-the-fly based on the human wearer's interactions. and (2) multi-view pre-enrollment where images of an instance to be tracked are stored in memory ahead of time. To address IT3DEgo, we first re-purpose methods from relevant areas, e.g., single object tracking (SOT) -- running SOT methods to track instances in 2D frames and lifting them to 3D using camera pose and depth. We also present a simple method that leverages pretrained segmentation and detection models to generate proposals from RGB frames and match proposals with enrolled instance images. Perhaps surprisingly, our extensive experiments show that our method (with no finetuning) significantly outperforms SOT-based approaches. We conclude by arguing that the problem of egocentric instance tracking is made easier by leveraging camera pose and using a 3D allocentric (world) coordinate representation.