In this paper, we focus on developing knowledge distillation (KD) for compact 3D detectors. We observe that off-the-shelf KD methods manifest their efficacy only when the teacher model and student counterpart share similar intermediate feature representations. This might explain why they are less effective in building extreme-compact 3D detectors where significant representation disparity arises due primarily to the intrinsic sparsity and irregularity in 3D point clouds. This paper presents a novel representation disparity-aware distillation (RDD) method to address the representation disparity issue and reduce performance gap between compact students and over-parameterized teachers. This is accomplished by building our RDD from an innovative perspective of information bottleneck (IB), which can effectively minimize the disparity of proposal region pairs from student and teacher in features and logits. Extensive experiments are performed to demonstrate the superiority of our RDD over existing KD methods. For example, our RDD increases mAP of CP-Voxel-S to 57.1% on nuScenes dataset, which even surpasses teacher performance while taking up only 42% FLOPs.
Robust 3D object detection in extreme weather and illumination conditions is a challenging task. While radars and thermal cameras are known for their resilience to these conditions, few studies have been conducted on radar-thermal fusion due to the lack of corresponding datasets. To address this gap, we first present a new multi-modal dataset called ThermRad, which includes a 3D LiDAR, a 4D radar, an RGB camera and a thermal camera. This dataset is unique because it includes data from all four sensors in extreme weather conditions, providing a valuable resource for future research in this area. To validate the robustness of 4D radars and thermal cameras for 3D object detection in challenging weather conditions, we propose a new multi-modal fusion method called RTDF-RCNN, which leverages the complementary strengths of 4D radars and thermal cameras to boost object detection performance. To further prove the effectiveness of our proposed framework, we re-implement state-of-the-art (SOTA) 3D detectors on our dataset as benchmarks for evaluation. Our method achieves significant enhancements in detecting cars, pedestrians, and cyclists, with improvements of over 7.98%, 24.27%, and 27.15%, respectively, while achieving comparable results to LiDAR-based approaches. Our contributions in both the ThermRad dataset and the new multi-modal fusion method provide a new approach to robust 3D object detection in adverse weather and illumination conditions. The ThermRad dataset will be released.
The task of Human-Object Interaction (HOI) detection is to detect humans and their interactions with surrounding objects, where transformer-based methods show dominant advances currently. However, these methods ignore the relationship among humans, objects, and interactions: 1) human features are more contributive than object ones to interaction prediction; 2) interactive information disturbs the detection of objects but helps human detection. In this paper, we propose a Human and Object Disentangling Network (HODN) to model the HOI relationships explicitly, where humans and objects are first detected by two disentangling decoders independently and then processed by an interaction decoder. Considering that human features are more contributive to interaction, we propose a Human-Guide Linking method to make sure the interaction decoder focuses on the human-centric regions with human features as the positional embeddings. To handle the opposite influences of interactions on humans and objects, we propose a Stop-Gradient Mechanism to stop interaction gradients from optimizing the object detection but to allow them to optimize the human detection. Our proposed method achieves competitive performance on both the V-COCO and the HICO-Det datasets. It can be combined with existing methods easily for state-of-the-art results.
This work focuses on the 3D reconstruction of non-rigid objects based on monocular RGB video sequences. Concretely, we aim at building high-fidelity models for generic object categories and casually captured scenes. To this end, we do not assume known root poses of objects, and do not utilize category-specific templates or dense pose priors. The key idea of our method, Root Pose Decomposition (RPD), is to maintain a per-frame root pose transformation, meanwhile building a dense field with local transformations to rectify the root pose. The optimization of local transformations is performed by point registration to the canonical space. We also adapt RPD to multi-object scenarios with object occlusions and individual differences. As a result, RPD allows non-rigid 3D reconstruction for complicated scenarios containing objects with large deformations, complex motion patterns, occlusions, and scale diversities of different individuals. Such a pipeline potentially scales to diverse sets of objects in the wild. We experimentally show that RPD surpasses state-of-the-art methods on the challenging DAVIS, OVIS, and AMA datasets.
Image composition refers to inserting a foreground object into a background image to obtain a composite image. In this work, we focus on generating plausible shadow for the inserted foreground object to make the composite image more realistic. To supplement the existing small-scale dataset DESOBA, we create a large-scale dataset called DESOBAv2 by using object-shadow detection and inpainting techniques. Specifically, we collect a large number of outdoor scene images with object-shadow pairs. Then, we use pretrained inpainting model to inpaint the shadow region, resulting in the deshadowed images. Based on real images and deshadowed images, we can construct pairs of synthetic composite images and ground-truth target images. Dataset is available at https://github.com/bcmi/Object-Shadow-Generation-Dataset-DESOBAv2.
Self-supervised methods have shown remarkable progress in learning high-level semantics and low-level temporal correspondence. Building on these results, we take one step further and explore the possibility of integrating these two features to enhance object-centric representations. Our preliminary experiments indicate that query slot attention can extract different semantic components from the RGB feature map, while random sampling based slot attention can exploit temporal correspondence cues between frames to assist instance identification. Motivated by this, we propose a novel semantic-aware masked slot attention on top of the fused semantic features and correspondence maps. It comprises two slot attention stages with a set of shared learnable Gaussian distributions. In the first stage, we use the mean vectors as slot initialization to decompose potential semantics and generate semantic segmentation masks through iterative attention. In the second stage, for each semantics, we randomly sample slots from the corresponding Gaussian distribution and perform masked feature aggregation within the semantic area to exploit temporal correspondence patterns for instance identification. We adopt semantic- and instance-level temporal consistency as self-supervision to encourage temporally coherent object-centric representations. Our model effectively identifies multiple object instances with semantic structure, reaching promising results on unsupervised video object discovery. Furthermore, we achieve state-of-the-art performance on dense label propagation tasks, demonstrating the potential for object-centric analysis. The code is released at https://github.com/shvdiwnkozbw/SMTC.
Rotation estimation of high precision from an RGB-D object observation is a huge challenge in 6D object pose estimation, due to the difficulty of learning in the non-linear space of SO(3). In this paper, we propose a novel rotation estimation network, termed as VI-Net, to make the task easier by decoupling the rotation as the combination of a viewpoint rotation and an in-plane rotation. More specifically, VI-Net bases the feature learning on the sphere with two individual branches for the estimates of two factorized rotations, where a V-Branch is employed to learn the viewpoint rotation via binary classification on the spherical signals, while another I-Branch is used to estimate the in-plane rotation by transforming the signals to view from the zenith direction. To process the spherical signals, a Spherical Feature Pyramid Network is constructed based on a novel design of SPAtial Spherical Convolution (SPA-SConv), which settles the boundary problem of spherical signals via feature padding and realizesviewpoint-equivariant feature extraction by symmetric convolutional operations. We apply the proposed VI-Net to the challenging task of category-level 6D object pose estimation for predicting the poses of unknown objects without available CAD models; experiments on the benchmarking datasets confirm the efficacy of our method, which outperforms the existing ones with a large margin in the regime of high precision.
Multi-object tracking (MOT) is a challenging vision task that aims to detect individual objects within a single frame and associate them across multiple frames. Recent MOT approaches can be categorized into two-stage tracking-by-detection (TBD) methods and one-stage joint detection and tracking (JDT) methods. Despite the success of these approaches, they also suffer from common problems, such as harmful global or local inconsistency, poor trade-off between robustness and model complexity, and lack of flexibility in different scenes within the same video. In this paper we propose a simple but robust framework that formulates object detection and association jointly as a consistent denoising diffusion process from paired noise boxes to paired ground-truth boxes. This novel progressive denoising diffusion strategy substantially augments the tracker's effectiveness, enabling it to discriminate between various objects. During the training stage, paired object boxes diffuse from paired ground-truth boxes to random distribution, and the model learns detection and tracking simultaneously by reversing this noising process. In inference, the model refines a set of paired randomly generated boxes to the detection and tracking results in a flexible one-step or multi-step denoising diffusion process. Extensive experiments on three widely used MOT benchmarks, including MOT17, MOT20, and Dancetrack, demonstrate that our approach achieves competitive performance compared to the current state-of-the-art methods.
The current popular methods for video object segmentation (VOS) implement feature matching through several hand-crafted modules that separately perform feature extraction and matching. However, the above hand-crafted designs empirically cause insufficient target interaction, thus limiting the dynamic target-aware feature learning in VOS. To tackle these limitations, this paper presents a scalable Simplified VOS (SimVOS) framework to perform joint feature extraction and matching by leveraging a single transformer backbone. Specifically, SimVOS employs a scalable ViT backbone for simultaneous feature extraction and matching between query and reference features. This design enables SimVOS to learn better target-ware features for accurate mask prediction. More importantly, SimVOS could directly apply well-pretrained ViT backbones (e.g., MAE) for VOS, which bridges the gap between VOS and large-scale self-supervised pre-training. To achieve a better performance-speed trade-off, we further explore within-frame attention and propose a new token refinement module to improve the running speed and save computational cost. Experimentally, our SimVOS achieves state-of-the-art results on popular video object segmentation benchmarks, i.e., DAVIS-2017 (88.0% J&F), DAVIS-2016 (92.9% J&F) and YouTube-VOS 2019 (84.2% J&F), without applying any synthetic video or BL30K pre-training used in previous VOS approaches.
This paper addresses the limitations of current datasets for 3D vision tasks in terms of accuracy, size, realism, and suitable imaging modalities for photometrically challenging objects. We propose a novel annotation and acquisition pipeline that enhances existing 3D perception and 6D object pose datasets. Our approach integrates robotic forward-kinematics, external infrared trackers, and improved calibration and annotation procedures. We present a multi-modal sensor rig, mounted on a robotic end-effector, and demonstrate how it is integrated into the creation of highly accurate datasets. Additionally, we introduce a freehand procedure for wider viewpoint coverage. Both approaches yield high-quality 3D data with accurate object and camera pose annotations. Our methods overcome the limitations of existing datasets and provide valuable resources for 3D vision research.
Vision Transformers (ViTs) have achieved remarkable success in computer vision tasks. However, their potential in rotation-sensitive scenarios has not been fully explored, and this limitation may be inherently attributed to the lack of spatial invariance in the data-forwarding process. In this study, we present a novel approach, termed Spatial Transform Decoupling (STD), providing a simple-yet-effective solution for oriented object detection with ViTs. Built upon stacked ViT blocks, STD utilizes separate network branches to predict the position, size, and angle of bounding boxes, effectively harnessing the spatial transform potential of ViTs in a divide-and-conquer fashion. Moreover, by aggregating cascaded activation masks (CAMs) computed upon the regressed parameters, STD gradually enhances features within regions of interest (RoIs), which complements the self-attention mechanism. Without bells and whistles, STD achieves state-of-the-art performance on the benchmark datasets including DOTA-v1.0 (82.24% mAP) and HRSC2016 (98.55% mAP), which demonstrates the effectiveness of the proposed method. Source code is available at https://github.com/yuhongtian17/Spatial-Transform-Decoupling.
Online multi-object tracking (MOT) plays a pivotal role in autonomous systems. The state-of-the-art approaches usually employ a tracking-by-detection method, and data association plays a critical role. This paper proposes a learning and graph-optimized (LEGO) modular tracker to improve data association performance in the existing literature. The proposed LEGO tracker integrates graph optimization and self-attention mechanisms, which efficiently formulate the association score map, facilitating the accurate and efficient matching of objects across time frames. To further enhance the state update process, the Kalman filter is added to ensure consistent tracking by incorporating temporal coherence in the object states. Our proposed method utilizing LiDAR alone has shown exceptional performance compared to other online tracking approaches, including LiDAR-based and LiDAR-camera fusion-based methods. LEGO ranked 1st at the time of submitting results to KITTI object tracking evaluation ranking board and remains 2nd at the time of submitting this paper, among all online trackers in the KITTI MOT benchmark for cars1
Camera-based 3D object detection in BEV (Bird's Eye View) space has drawn great attention over the past few years. Dense detectors typically follow a two-stage pipeline by first constructing a dense BEV feature and then performing object detection in BEV space, which suffers from complex view transformations and high computation cost. On the other side, sparse detectors follow a query-based paradigm without explicit dense BEV feature construction, but achieve worse performance than the dense counterparts. In this paper, we find that the key to mitigate this performance gap is the adaptability of the detector in both BEV and image space. To achieve this goal, we propose SparseBEV, a fully sparse 3D object detector that outperforms the dense counterparts. SparseBEV contains three key designs, which are (1) scale-adaptive self attention to aggregate features with adaptive receptive field in BEV space, (2) adaptive spatio-temporal sampling to generate sampling locations under the guidance of queries, and (3) adaptive mixing to decode the sampled features with dynamic weights from the queries. On the test split of nuScenes, SparseBEV achieves the state-of-the-art performance of 67.5 NDS. On the val split, SparseBEV achieves 55.8 NDS while maintaining a real-time inference speed of 23.5 FPS. Code is available at https://github.com/MCG-NJU/SparseBEV.
Resource-constrained perception systems such as edge computing and vision-for-robotics require vision models to be both accurate and lightweight in computation and memory usage. While knowledge distillation is a proven strategy to enhance the performance of lightweight classification models, its application to structured outputs like object detection and instance segmentation remains a complicated task, due to the variability in outputs and complex internal network modules involved in the distillation process. In this paper, we propose a simple yet surprisingly effective sequential approach to knowledge distillation that progressively transfers the knowledge of a set of teacher detectors to a given lightweight student. To distill knowledge from a highly accurate but complex teacher model, we construct a sequence of teachers to help the student gradually adapt. Our progressive strategy can be easily combined with existing detection distillation mechanisms to consistently maximize student performance in various settings. To the best of our knowledge, we are the first to successfully distill knowledge from Transformer-based teacher detectors to convolution-based students, and unprecedentedly boost the performance of ResNet-50 based RetinaNet from 36.5% to 42.0% AP and Mask R-CNN from 38.2% to 42.5% AP on the MS COCO benchmark.
In this paper, we demonstrate that the concept of Semantic Consistency and the ensuing method of Knowledge-Aware Re-Optimization can be adapted for the problem of object detection in intricate traffic scenes. Furthermore, we introduce a novel method for extracting a knowledge graph from a dataset of images provided with instance-level annotations, and integrate this new knowledge graph with the existing semantic consistency model. Combining both this novel hybrid knowledge graph and the preexisting methods of frequency analysis and external knowledge graph as sources for semantic information, we investigate the effectiveness of knowledge-aware re-optimization on the Faster-RCNN and DETR object detection models. We find that limited but consistent improvements in precision and or recall can be achieved using this method for all combinations of model and method studied.
Camouflaged object detection (COD) aims to accurately detect objects hidden in the surrounding environment. However, the existing COD methods mainly locate camouflaged objects in the RGB domain, their performance has not been fully exploited in many challenging scenarios. Considering that the features of the camouflaged object and the background are more discriminative in the frequency domain, we propose a novel learnable and separable frequency perception mechanism driven by the semantic hierarchy in the frequency domain. Our entire network adopts a two-stage model, including a frequency-guided coarse localization stage and a detail-preserving fine localization stage. With the multi-level features extracted by the backbone, we design a flexible frequency perception module based on octave convolution for coarse positioning. Then, we design the correction fusion module to step-by-step integrate the high-level features through the prior-guided correction and cross-layer feature channel association, and finally combine them with the shallow features to achieve the detailed correction of the camouflaged objects. Compared with the currently existing models, our proposed method achieves competitive performance in three popular benchmark datasets both qualitatively and quantitatively.
Recently there has been a growing interest in category-level object pose and size estimation, and prevailing methods commonly rely on single view RGB-D images. However, one disadvantage of such methods is that they require accurate depth maps which cannot be produced by consumer-grade sensors. Furthermore, many practical real-world situations involve a moving camera that continuously observes its surroundings, and the temporal information of the input video streams is simply overlooked by single-view methods. We propose a novel solution that makes use of RGB video streams. Our framework consists of three modules: a scale-aware monocular dense SLAM solution, a lightweight object pose predictor, and an object-level pose graph optimizer. The SLAM module utilizes a video stream and additional scale-sensitive readings to estimate camera poses and metric depth. The object pose predictor then generates canonical object representations from RGB images. The object pose is estimated through geometric registration of these canonical object representations with estimated object depth points. All per-view estimates finally undergo optimization within a pose graph, culminating in the output of robust and accurate canonical object poses. Our experimental results demonstrate that when utilizing public dataset sequences with high-quality depth information, the proposed method exhibits comparable performance to state-of-the-art RGB-D methods. We also collect and evaluate on new datasets containing depth maps of varying quality to further quantitatively benchmark the proposed method alongside previous RGB-D based methods. We demonstrate a significant advantage in scenarios where depth input is absent or the quality of depth sensing is limited.
Single-image 3D reconstruction is a research challenge focused on predicting 3D object shapes from single-view images. This task requires significant data acquisition to predict both visible and occluded portions of the shape. Furthermore, learning-based methods face the difficulty of creating a comprehensive training dataset for all possible classes. To this end, we propose a continual learning-based 3D reconstruction method where our goal is to design a model using Variational Priors that can still reconstruct the previously seen classes reasonably even after training on new classes. Variational Priors represent abstract shapes and combat forgetting, whereas saliency maps preserve object attributes with less memory usage. This is vital due to resource constraints in storing extensive training data. Additionally, we introduce saliency map-based experience replay to capture global and distinct object features. Thorough experiments show competitive results compared to established methods, both quantitatively and qualitatively.
Tracking 3D objects is an important task in autonomous driving. Classical Kalman Filtering based methods are still the most popular solutions. However, these methods require handcrafted designs in motion modeling and can not benefit from the growing data amounts. In this paper, Box Only Transformer Tracker (BOTT) is proposed to learn to link 3D boxes of the same object from the different frames, by taking all the 3D boxes in a time window as input. Specifically, transformer self-attention is applied to exchange information between all the boxes to learn global-informative box embeddings. The similarity between these learned embeddings can be used to link the boxes of the same object. BOTT can be used for both online and offline tracking modes seamlessly. Its simplicity enables us to significantly reduce engineering efforts required by traditional Kalman Filtering based methods. Experiments show BOTT achieves competitive performance on two largest 3D MOT benchmarks: 69.9 and 66.7 AMOTA on nuScenes validation and test splits, respectively, 56.45 and 59.57 MOTA L2 on Waymo Open Dataset validation and test splits, respectively. This work suggests that tracking 3D objects by learning features directly from 3D boxes using transformers is a simple yet effective way.
Query-based object detectors directly decode image features into object instances with a set of learnable queries. These query vectors are progressively refined to stable meaningful representations through a sequence of decoder layers, and then used to directly predict object locations and categories with simple FFN heads. In this paper, we present a new query-based object detector (DEQDet) by designing a deep equilibrium decoder. Our DEQ decoder models the query vector refinement as the fixed point solving of an {implicit} layer and is equivalent to applying {infinite} steps of refinement. To be more specific to object decoding, we use a two-step unrolled equilibrium equation to explicitly capture the query vector refinement. Accordingly, we are able to incorporate refinement awareness into the DEQ training with the inexact gradient back-propagation (RAG). In addition, to stabilize the training of our DEQDet and improve its generalization ability, we devise the deep supervision scheme on the optimization path of DEQ with refinement-aware perturbation~(RAP). Our experiments demonstrate DEQDet converges faster, consumes less memory, and achieves better results than the baseline counterpart (AdaMixer). In particular, our DEQDet with ResNet50 backbone and 300 queries achieves the $49.5$ mAP and $33.0$ AP$_s$ on the MS COCO benchmark under $2\times$ training scheme (24 epochs).