Fine-grained urban flow inference (FUFI) is crucial for traffic management, as it infers high-resolution urban flow maps from coarse-grained observations. Existing FUFI methods typically focus on a single city and rely on comprehensive training with large-scale datasets to achieve precise inferences. However, data availability in developing cities may be limited, posing challenges to the development of well-performing models. To address this issue, we propose cross-city fine-grained urban flow inference, which aims to transfer spatio-temporal knowledge from data-rich cities to data-scarce areas using meta-transfer learning. This paper devises a Spatio-Temporal Deviation Alignment (STDA) framework to mitigate spatio-temporal distribution deviations and urban structural deviations between multiple source cities and the target city. Furthermore, STDA presents a cross-city normalization method that adaptively combines batch and instance normalization to maintain consistency between city-variant and city-invariant features. Besides, we design an urban structure alignment module to align spatial topological differences across cities. STDA effectively reduces distribution and structural deviations among different datasets while avoiding negative transfer. Extensive experiments conducted on three real-world datasets demonstrate that STDA consistently outperforms state-of-the-art baselines.

LiDAR has become one of the primary 3D object detection sensors in autonomous driving. However, due to the inherent sparsity of point clouds, certain objects exhibit structure incompleteness in occluded and distant areas, which hampers the accurate perception of objects in 3D space. To tackle this challenge, we propose Semantic and Structure Completion Network (S2CNet) for 3D object detection. Concretely, we design the Semantic Completion (SeC) module to generate semantic features in Bird’s-Eye-View (BEV) space, utilizing a teacher-student paradigm. Notably, we adopt a coarse-to-fine guidance strategy to encourage student network to generate semantic features specifically within foreground regions. This ensures that the student network focuses on the generation of foreground object features. Besides, we introduce an attention-based module to adaptively fuse the generated features and raw features. SeC module faces particular limitation when dealing with objects containing only a few points, in such case, the network is prone to generating low quality proposals with inaccurate localization. Complementary to SeC module, we introduce the Structure Completion (StC) module, in which a group of structural proposals are obtained by traversing most structures in a structure-guided manner, and thus at least one proposal with ground truth similar structure can be guaranteed. Extensive experiments on the KITTI and nuScenes benchmarks demonstrate the effectiveness of our method, especially for the hard setting objects with fewer points.

Accurate pedestrian classification and localization has garnered significant attention due to their extensive applications in various multimedia applications such as security monitoring, autonomous driving, and more. We have observed that the commonly employed Intersection over Union (IoU) metric in many pedestrian detectors is susceptible to an inconsistent GT-Proposal assignment issue. This issue arises when spatially adjacent proposals, which have highly similar features, are assigned to distinct ground-truth boxes, leading to confusion during the training process and an increased number of false positives during inference. To address this challenge, our work presents a novel algorithm namedDirectionalAssignmentStrategy (DAS). Firstly, in conjunction with depth distribution, our approach transforms the assignment metric from a two-dimensional (2D) view into a three-dimensional (3D) space, enabling the optimization of the regression head under the constraint of depth direction. Secondly, in contrast to the conventional IoU-basedone-to-oneassignment of one proposal to one ground-truth box, our method aims to establish a more reasoned matching between sets of proposals and ground-truth boxes. By doing so, the detector is less reliant on the setting of a specific threshold. Leveraging this strategy as a plug-in module within state-of-the-art pedestrian detectors, we demonstrate a notable improvement in performance.

Place Recognition enables the estimation of a globally consistent map and trajectory by providing non-local constraints in Simultaneous Localisation and Mapping (SLAM). This paper presents Locus, a novel place recognition method using 3D LiDAR point clouds in large-scale environments. We propose a method for extracting and encoding topological and temporal information related to components in a scene and demonstrate how the inclusion of this auxiliary information in place description leads to more robust and discriminative scene representations. Second-order pooling along with a non- linear transform is used to aggregate these multi-level features to generate a fixed-length global descriptor, which is invariant to the permutation of input features. The proposed method outperforms state-of-the-art methods on the KITTI dataset. Furthermore, Locus is demonstrated to be robust across several challenging situations such as occlusions and viewpoint changes in 3D LiDAR point clouds. The open-source implementation is available at: https://github.com/csiro-robotics/locus.

Visual relationship detection, as a challenging task used to find and distinguish interactions between object-pairs in one image, has received much attention recently. In this work, we devise a unified visual relationship detection framework with two types of correlation exploitation to address the combination explosion problem in the object-pairs proposing stage and the non-exclusive label problem in the predicate recognition stage. In the object-pairs proposing stage, with the exploitation of relative location correlation between two objects in one pair, one location-embedded rating module (LRM) is developed to effectively select plausible proposals. In the predicate recognition stage, one label-correlation graph module (LGM) is introduced to measure the implicit semantic correlation among predicates; and then assign discrete distributed labels to predicates to improve the precision of top-n recall. Experiments on the two widely used VRD and VG datasets show that our proposed method outperforms current state-of-the-art methods.

Due to a wide range of various application scenes, robust crowd counting is still quite difficult and the performance is far from being satisfied. In this paper, we propose a novel robust crowd counting method by introducing a weakly supervised crowd-wise attention network. The proposed work improves the counting accuracy and robustness by: i) Weakly-supervised crowd segmentation. With a generated segmentation label using motion-guided region-growth, both the appearance feature of one-labeled image and motion features abstracted from its adjacent unlabeled frames, are combined to implement weakly supervised crowd region segmentation, with which active crowd region can be finely perceived from different background disturbances. ii) More accurate spatial attention. We generate a spatial attention map based on the active crowd segmentation, which is used to reweigh the appearance feature to achieve attention-based density estimation. Evaluation of the widely used World Expo' 10 dataset shows that the proposed work can achieve state-of-the-art performance on both accuracy and robustness.

Traditional static-image based crowd counting methods work well on public datasets. However, due to the complexity and variability of real-world scenarios, their performance tends to drop dramatically in practice. Aiming to solve the robust problem of crowd counting, we propose to use a co-attention mechanism to extract correlation features lying behind adjacent video frames which can enhance the distinguish-ability between background and foreground. Also, we combine co-attention with spatial attention and multi-scale self-attention. Three different and complementary attention-based modules jointly reinforce the robustness of the counting model. Experiments on two widely used video crowd datasets demonstrate the effectiveness of the proposed approach.

Scale variation of pedestrians in a crowd image presents a significant challenge for vision-based people counting systems. Such variations are mainly caused by perspective-related distortions due to the camera pose relative to the ground plane. Following the density-based counting paradigm, we postulate that generating density values adaptive to object scales plays a critical role in the accuracy of the final counting results. Motivated by this, we distill the underlying information from depth cues to obtain scale-aware representations that can respond to object scales considering the fact that the scale is inversely proportional to the object depth. Specifically, we propose a depth embedding module as add-ons into existing networks. This module exploits essential depth cues to spatially re-calibrate the magnitude of the original features. In this way, the objects, although in the same class, will attain distinct representations according to their scales, which directly benefits the estimation of scale-aware density values. We conduct a comprehensive analysis of the effects of the depth embedding module and validate that exploiting depth cues to perceive object scale variations in convolutional neural networks improves crowd counting performances. Our experiments demonstrate the effectiveness of the proposed approach on four popular benchmark datasets.

Crowd counting is a challenging task in the presence of drastic scale variations, the clutter background, and severe occlusions, etc. Existing CNN-based counting methods tackle these challenges mainly by fusing either multi-scale or multi-context features to generate robust representations. In this paper, we propose to address these issues by leveraging the heterogeneous attributes compounded in the density map. We identify three geometric/semantic/numeric attributes essentially important to the density estimation, and demonstrate how to effectively utilize these heterogeneous attributes to assist the crowd counting by formulating them into multiple auxiliary tasks. With the multi-fold regularization effects induced by the auxiliary tasks, the backbone CNN model is driven to embed desired properties explicitly and thus gains robust representations towards more accurate density estimation. Extensive experiments on three challenging crowd counting datasets have demonstrated the effectiveness of the proposed approach.

We present a novel deep learning framework for crowd counting by learning a perspective-embedded deconvolution network. Perspective is an inherent property of most surveillance scenes. Unlike the traditional approaches that exploit the perspective as a separate normalization, we propose to fuse the perspective into a deconvolution network, aiming to obtain a robust, accurate and consistent crowd density map. Through layer-wise fusion, we merge perspective maps at different resolutions into the deconvolution network. With the injection of perspective, our network is driven to learn to combine the underlying scene geometric constraints adaptively, thus enabling an accurate interpretation from high-level feature maps to the pixel-wise crowd density map. In addition, our network allows generating density map for arbitrary-sized input in an end-to-end fashion. The proposed method achieves competitive result on the WorldExpo2010 crowd dataset.

Cloud detection plays an important role in total-sky images based solar forecasting and has received more attention in recent years. Accurate cloud detection for complicated total-sky images is especially changeling due to the low contrast and vague boundaries between cloud and sky regions. Unlike the existing cloud detection method without any preprocessing, one novel decorrelation-stretch (DS) based method is proposed in this work, where the total-sky images are preprocessed using the DS algorithm firstly. With this enhancement, color feature disparity of cloud and sky can be intensified notably, and then a more accurate threshold can be obtained by applying the Minimum Cross Entropy (MCE) to the preprocessed image. Experimental results demonstrated the proposed scheme achieves better performance than the existing cloud detection methods on total-sky images, especially for images with low contrast or vague boundaries between cloud and sky regions.