Abstract:Task-specific data-fusion networks have marked considerable achievements in urban scene parsing. Among these networks, our recently proposed RoadFormer successfully extracts heterogeneous features from RGB images and surface normal maps and fuses these features through attention mechanisms, demonstrating compelling efficacy in RGB-Normal road scene parsing. However, its performance significantly deteriorates when handling other types/sources of data or performing more universal, all-category scene parsing tasks. To overcome these limitations, this study introduces RoadFormer+, an efficient, robust, and adaptable model capable of effectively fusing RGB-X data, where ``X'', represents additional types/modalities of data such as depth, thermal, surface normal, and polarization. Specifically, we propose a novel hybrid feature decoupling encoder to extract heterogeneous features and decouple them into global and local components. These decoupled features are then fused through a dual-branch multi-scale heterogeneous feature fusion block, which employs parallel Transformer attentions and convolutional neural network modules to merge multi-scale features across different scales and receptive fields. The fused features are subsequently fed into a decoder to generate the final semantic predictions. Notably, our proposed RoadFormer+ ranks first on the KITTI Road benchmark and achieves state-of-the-art performance in mean intersection over union on the Cityscapes, MFNet, FMB, and ZJU datasets. Moreover, it reduces the number of learnable parameters by 65\% compared to RoadFormer. Our source code will be publicly available at mias.group/RoadFormerPlus.
Abstract:Unmanned vehicles usually rely on Global Positioning System (GPS) and Light Detection and Ranging (LiDAR) sensors to achieve high-precision localization results for navigation purpose. However, this combination with their associated costs and infrastructure demands, poses challenges for widespread adoption in mass-market applications. In this paper, we aim to use only a monocular camera to achieve comparable onboard localization performance by tracking deep-learning visual features on a LiDAR-enhanced visual prior map. Experiments show that the proposed algorithm can provide centimeter-level global positioning results with scale, which is effortlessly integrated and favorable for low-cost robot system deployment in real-world applications.
Abstract:Despite the remarkable developments of recent large models in Embodied Artificial Intelligence (E-AI), their integration into robotics is hampered by their excessive parameter sizes and computational demands. Towards the Vision-and-Language Navigation (VLN) task, a core task in E-AI, this paper reveals the great potential of using knowledge distillation for obtaining lightweight student models by proposing a Meta-Ability Guided Interactive Chain-of-distillation (MAGIC) method. Specifically, a Meta-Ability Knowledge Distillation (MAKD) framework is proposed for decoupling and refining the necessary meta-abilities of VLN agents. A Meta-Knowledge Randomization Weighting (MKRW) and a Meta-Knowledge Transferable Determination (MKTD) module are incorporated to dynamically adjust aggregation weights at the meta-ability and sample levels, respectively. Move beyond the traditional one-step unidirectional distillation, an Interactive Chain-of-Distillation (ICoD) learning strategy is proposed to allow students to give feedback to teachers, forming a new multi-step teacher-student co-evolution pipeline. Remarkably, on the R2R test unseen public leaderboard, our smallest model, MAGIC-S, with only 5% (11M) of the teacher's size, outperforms all previous methods under the same training data. Additionally, our largest model, MAGIC-L, surpasses the previous state-of-the-art by 5.84% in SPL and 3.18% in SR. Furthermore, a new dataset was collected and annotated from our living environments, where MAGIC-S demonstrated superior performance and real-time efficiency. Our code is publicly available on https://github.com/CrystalSixone/VLN-MAGIC.
Abstract:Visual anomaly detection is vital in real-world applications, such as industrial defect detection and medical diagnosis. However, most existing methods focus on local structural anomalies and fail to detect higher-level functional anomalies under logical conditions. Although recent studies have explored logical anomaly detection, they can only address simple anomalies like missing or addition and show poor generalizability due to being heavily data-driven. To fill this gap, we propose SAM-LAD, a zero-shot, plug-and-play framework for logical anomaly detection in any scene. First, we obtain a query image's feature map using a pre-trained backbone. Simultaneously, we retrieve the reference images and their corresponding feature maps via the nearest neighbor search of the query image. Then, we introduce the Segment Anything Model (SAM) to obtain object masks of the query and reference images. Each object mask is multiplied with the entire image's feature map to obtain object feature maps. Next, an Object Matching Model (OMM) is proposed to match objects in the query and reference images. To facilitate object matching, we further propose a Dynamic Channel Graph Attention (DCGA) module, treating each object as a keypoint and converting its feature maps into feature vectors. Finally, based on the object matching relations, an Anomaly Measurement Model (AMM) is proposed to detect objects with logical anomalies. Structural anomalies in the objects can also be detected. We validate our proposed SAM-LAD using various benchmarks, including industrial datasets (MVTec Loco AD, MVTec AD), and the logical dataset (DigitAnatomy). Extensive experimental results demonstrate that SAM-LAD outperforms existing SoTA methods, particularly in detecting logical anomalies.
Abstract:Motion diffusion models have recently proven successful for text-driven human motion generation. Despite their excellent generation performance, they are challenging to infer in real time due to the multi-step sampling mechanism that involves tens or hundreds of repeat function evaluation iterations. To this end, we investigate a motion latent consistency Training (MLCT) for motion generation to alleviate the computation and time consumption during iteration inference. It applies diffusion pipelines to low-dimensional motion latent spaces to mitigate the computational burden of each function evaluation. Explaining the diffusion process with probabilistic flow ordinary differential equation (PF-ODE) theory, the MLCT allows extremely few steps infer between the prior distribution to the motion latent representation distribution via maintaining consistency of the outputs over the trajectory of PF-ODE. Especially, we introduce a quantization constraint to optimize motion latent representations that are bounded, regular, and well-reconstructed compared to traditional variational constraints. Furthermore, we propose a conditional PF-ODE trajectory simulation method, which improves the conditional generation performance with minimal additional training costs. Extensive experiments on two human motion generation benchmarks show that the proposed model achieves state-of-the-art performance with less than 10\% time cost.
Abstract:In the pursuit of robust and generalizable environment perception and language understanding, the ubiquitous challenge of dataset bias continues to plague vision-and-language navigation (VLN) agents, hindering their performance in unseen environments. This paper introduces the generalized cross-modal causal transformer (GOAT), a pioneering solution rooted in the paradigm of causal inference. By delving into both observable and unobservable confounders within vision, language, and history, we propose the back-door and front-door adjustment causal learning (BACL and FACL) modules to promote unbiased learning by comprehensively mitigating potential spurious correlations. Additionally, to capture global confounder features, we propose a cross-modal feature pooling (CFP) module supervised by contrastive learning, which is also shown to be effective in improving cross-modal representations during pre-training. Extensive experiments across multiple VLN datasets (R2R, REVERIE, RxR, and SOON) underscore the superiority of our proposed method over previous state-of-the-art approaches. Code is available at https://github.com/CrystalSixone/VLN-GOAT.
Abstract:Vision-and-Language Navigation (VLN) has gained significant research interest in recent years due to its potential applications in real-world scenarios. However, existing VLN methods struggle with the issue of spurious associations, resulting in poor generalization with a significant performance gap between seen and unseen environments. In this paper, we tackle this challenge by proposing a unified framework CausalVLN based on the causal learning paradigm to train a robust navigator capable of learning unbiased feature representations. Specifically, we establish reasonable assumptions about confounders for vision and language in VLN using the structured causal model (SCM). Building upon this, we propose an iterative backdoor-based representation learning (IBRL) method that allows for the adaptive and effective intervention on confounders. Furthermore, we introduce the visual and linguistic backdoor causal encoders to enable unbiased feature expression for multi-modalities during training and validation, enhancing the agent's capability to generalize across different environments. Experiments on three VLN datasets (R2R, RxR, and REVERIE) showcase the superiority of our proposed method over previous state-of-the-art approaches. Moreover, detailed visualization analysis demonstrates the effectiveness of CausalVLN in significantly narrowing down the performance gap between seen and unseen environments, underscoring its strong generalization capability.
Abstract:Most of existing category-level object pose estimation methods devote to learning the object category information from point cloud modality. However, the scale of 3D datasets is limited due to the high cost of 3D data collection and annotation. Consequently, the category features extracted from these limited point cloud samples may not be comprehensive. This motivates us to investigate whether we can draw on knowledge of other modalities to obtain category information. Inspired by this motivation, we propose CLIPose, a novel 6D pose framework that employs the pre-trained vision-language model to develop better learning of object category information, which can fully leverage abundant semantic knowledge in image and text modalities. To make the 3D encoder learn category-specific features more efficiently, we align representations of three modalities in feature space via multi-modal contrastive learning. In addition to exploiting the pre-trained knowledge of the CLIP's model, we also expect it to be more sensitive with pose parameters. Therefore, we introduce a prompt tuning approach to fine-tune image encoder while we incorporate rotations and translations information in the text descriptions. CLIPose achieves state-of-the-art performance on two mainstream benchmark datasets, REAL275 and CAMERA25, and runs in real-time during inference (40FPS).
Abstract:Estimating the 6D object pose is an essential task in many applications. Due to the lack of depth information, existing RGB-based methods are sensitive to occlusion and illumination changes. How to extract and utilize the geometry features in depth information is crucial to achieve accurate predictions. To this end, we propose TransPose, a novel 6D pose framework that exploits Transformer Encoder with geometry-aware module to develop better learning of point cloud feature representations. Specifically, we first uniformly sample point cloud and extract local geometry features with the designed local feature extractor base on graph convolution network. To improve robustness to occlusion, we adopt Transformer to perform the exchange of global information, making each local feature contains global information. Finally, we introduce geometry-aware module in Transformer Encoder, which to form an effective constrain for point cloud feature learning and makes the global information exchange more tightly coupled with point cloud tasks. Extensive experiments indicate the effectiveness of TransPose, our pose estimation pipeline achieves competitive results on three benchmark datasets.
Abstract:We propose InstructDET, a data-centric method for referring object detection (ROD) that localizes target objects based on user instructions. While deriving from referring expressions (REC), the instructions we leverage are greatly diversified to encompass common user intentions related to object detection. For one image, we produce tremendous instructions that refer to every single object and different combinations of multiple objects. Each instruction and its corresponding object bounding boxes (bbxs) constitute one training data pair. In order to encompass common detection expressions, we involve emerging vision-language model (VLM) and large language model (LLM) to generate instructions guided by text prompts and object bbxs, as the generalizations of foundation models are effective to produce human-like expressions (e.g., describing object property, category, and relationship). We name our constructed dataset as InDET. It contains images, bbxs and generalized instructions that are from foundation models. Our InDET is developed from existing REC datasets and object detection datasets, with the expanding potential that any image with object bbxs can be incorporated through using our InstructDET method. By using our InDET dataset, we show that a conventional ROD model surpasses existing methods on standard REC datasets and our InDET test set. Our data-centric method InstructDET, with automatic data expansion by leveraging foundation models, directs a promising field that ROD can be greatly diversified to execute common object detection instructions.