Vision-based Bird's Eye View (BEV) representation is an emerging perception formulation for autonomous driving. The core challenge is to construct BEV space with multi-camera features, which is a one-to-many ill-posed problem. Diving into all previous BEV representation generation methods, we found that most of them fall into two types: modeling depths in image views or modeling heights in the BEV space, mostly in an implicit way. In this work, we propose to explicitly model heights in the BEV space, which needs no extra data like LiDAR and can fit arbitrary camera rigs and types compared to modeling depths. Theoretically, we give proof of the equivalence between height-based methods and depth-based methods. Considering the equivalence and some advantages of modeling heights, we propose HeightFormer, which models heights and uncertainties in a self-recursive way. Without any extra data, the proposed HeightFormer could estimate heights in BEV accurately. Benchmark results show that the performance of HeightFormer achieves SOTA compared with those camera-only methods.

Modern methods mainly regard lane detection as a problem of pixel-wise segmentation, which is struggling to address the problems of efficiency and challenging scenarios like severe occlusions and extreme lighting conditions. Inspired by human perception, the recognition of lanes under severe occlusions and extreme lighting conditions is mainly based on contextual and global information. Motivated by this observation, we propose a novel, simple, yet effective formulation aiming at ultra fast speed and the problem of challenging scenarios. Specifically, we treat the process of lane detection as an anchor-driven ordinal classification problem using global features. First, we represent lanes with sparse coordinates on a series of hybrid (row and column) anchors. With the help of the anchor-driven representation, we then reformulate the lane detection task as an ordinal classification problem to get the coordinates of lanes. Our method could significantly reduce the computational cost with the anchor-driven representation. Using the large receptive field property of the ordinal classification formulation, we could also handle challenging scenarios. Extensive experiments on four lane detection datasets show that our method could achieve state-of-the-art performance in terms of both speed and accuracy. A lightweight version could even achieve 300+ frames per second(FPS). Our code is at https://github.com/cfzd/Ultra-Fast-Lane-Detection-v2.

Gait recognition aims at identifying the pedestrians at a long distance by their biometric gait patterns. It is inherently challenging due to the various covariates and the properties of silhouettes (textureless and colorless), which result in two kinds of pair-wise hard samples: the same pedestrian could have distinct silhouettes (intra-class diversity) and different pedestrians could have similar silhouettes (inter-class similarity). In this work, we propose to solve the hard sample issue with a Memory-augmented Progressive Learning network (GaitMPL), including Dynamic Reweighting Progressive Learning module (DRPL) and Global Structure-Aligned Memory bank (GSAM). Specifically, DRPL reduces the learning difficulty of hard samples by easy-to-hard progressive learning. GSAM further augments DRPL with a structure-aligned memory mechanism, which maintains and models the feature distribution of each ID. Experiments on two commonly used datasets, CASIA-B and OU-MVLP, demonstrate the effectiveness of GaitMPL. On CASIA-B, we achieve the state-of-the-art performance, i.e., 88.0% on the most challenging condition (Clothing) and 93.3% on the average condition, which outperforms the other methods by at least 3.8% and 1.4%, respectively. Code will be available at https://github.com/WhiteDOU/GaitMPL https://github.com/WhiteDOU/GaitMPL.

Bird’s eye view (BEV) representation is a new perception formulation for autonomous driving, which is based on spatial fusion. Further, temporal fusion is also introduced in BEV representation and gains great success. In this work, we propose a new method that unifies both spatial and temporal fusion and merges them into a unified mathematical formulation. The unified fusion could not only provide a new perspective on BEV fusion but also brings new capabilities. With the proposed unified spatial-temporal fusion, our method could support long-range fusion, which is hard to achieve in conventional BEV methods. Moreover, the BEV fusion in our work is temporal-adaptive and the weights of temporal fusion are learnable. In contrast, conventional methods mainly use fixed and equal weights for temporal fusion. Besides, the proposed unified fusion could avoid information lost in conventional BEV fusion methods and make full use of features. Extensive experiments and ablation studies on the NuScenes dataset show the effectiveness of the proposed method and our method gains the state-of-the-art performance in the map and vehicle segmentation task.

Gait recognition has a rapid development in recent years. However, current gait recognition focuses primarily on ideal laboratory scenes, leaving the gait in the wild unexplored. One of the main reasons is the difficulty of collecting in-the-wild gait datasets, which must ensure diversity of both intrinsic and extrinsic human gait factors. To remedy this problem, we propose to construct a large-scale gait dataset with the help of controllable computer simulation. In detail, to diversify the intrinsic factors of gait, we generate numerous characters with diverse attributes and associate them with various types of walking styles. To diversify the extrinsic factors of gait, we build a complicated scene with a dense camera layout. Then we design an automatic generation toolkit under Unity3D for simulating the walking scenarios and capturing the gait data. As a result, we obtain a dataset simulating towards the in-the-wild scenario, called VersatileGait, which has more than one million silhouette sequences of 10,000 subjects with diverse scenarios. VersatileGait possesses several nice properties, including huge dataset size, diverse pedestrian attributes, complicated camera layout, high-quality annotations, small domain gap with the real one, good scalability for new demands, and no privacy issues. By conducting a series of experiments, we first explore the effects of different factors on gait recognition. We further illustrate the effectiveness of using our dataset to pre-train models, which obtain considerable performance gain on CASIA-B, OU-MVLP, and CASIA-E. Besides, we show the great potential of the fine-grained labels other than the ID label in improving the efficiency and effectiveness of models. Our dataset and its corresponding generation toolkit are available at https://github.com/peterzpy/VersatileGait.

Monocular 3D object detection has attracted great attention for its advantages in simplicity and cost. Due to the ill-posed 2D to 3D mapping essence from the monocular imaging process, monocular 3D object detection suffers from inaccurate depth estimation and thus has poor 3D detection results. To alleviate this problem, we propose to introduce the ground plane as a prior in the monocular 3d object detection. The ground plane prior serves as an additional geometric condition to the ill-posed mapping and an extra source in depth estimation. In this way, we can get a more accurate depth estimation from the ground. Meanwhile, to take full advantage of the ground plane prior, we propose a depth-align training strategy and a precise two-stage depth inference method tailored for the ground plane prior. It is worth noting that the introduced ground plane prior requires no extra data sources like LiDAR, stereo images, and depth information. Extensive experiments on the KITTI benchmark show that our method could achieve state-of-the-art results compared with other methods while maintaining a very fast speed. Our code, models, and training logs are available at https://github.com/cfzd/MonoGround.

Dynamic routing networks, aimed at finding the best routing paths in the networks, have achieved significant improvements to neural networks in terms of accuracy and efficiency. In this paper, we see dynamic routing networks in a fresh light, formulating a routing method as a mapping from a sample space to a routing space. From the perspective of space mapping, prevalent methods of dynamic routing did not take into account how inference paths would be distributed in the routing space. Thus, we propose a novel method, termed CoDiNet, to model the relationship between a sample space and a routing space by regularizing the distribution of routing paths with the properties of consistency and diversity. Specifically, samples with similar semantics should be mapped into the same area in routing space, while those with dissimilar semantics should be mapped into different areas. Moreover, we design a customizable dynamic routing module, which can strike a balance between accuracy and efficiency. When deployed upon ResNet models, our method achieves higher performance and effectively reduces average computational cost on four widely used datasets.

Many CNN-based segmentation methods have been applied in lane marking detection recently and gain excellent success for a strong ability in modeling semantic information. Although the accuracy of lane line prediction is getting better and better, lane markings' localization ability is relatively weak, especially when the lane marking point is remote. Traditional lane detection methods usually utilize highly specialized handcrafted features and carefully designed postprocessing to detect the lanes. However, these methods are based on strong assumptions and, thus, are prone to scalability. In this work, we propose a novel multitask method that: 1) integrates the ability to model semantic information of CNN and the strong localization ability provided by handcrafted features and 2) predicts the position of vanishing line. A novel lane fitting method based on vanishing line prediction is also proposed for sharp curves and nonflat road in this article. By integrating segmentation, specialized handcrafted features, and fitting, the accuracy of location and the convergence speed of networks are improved. Extensive experimental results on four-lane marking detection data sets show that our method achieves state-of-the-art performance.

Attention mechanism, especially channel attention, has gained great success in the computer vision field. Many works focus on how to design efficient channel attention mechanisms while ignoring a fundamental problem, i.e., channel attention mechanism uses scalar to represent channel, which is difficult due to massive information loss. In this work, we start from a different view and regard the channel representation problem as a compression process using frequency analysis. Based on the frequency analysis, we mathematically prove that the conventional global average pooling is a special case of the feature decomposition in the frequency domain. With the proof, we naturally generalize the compression of the channel attention mechanism in the frequency domain and propose our method with multi-spectral channel attention, termed as FcaNet. FcaNet is simple but effective. We can change a few lines of code in the calculation to implement our method within existing channel attention methods. Moreover, the proposed method achieves state-of-the-art results compared with other channel attention methods on image classification, object detection, and instance segmentation tasks. Our method could consistently outperform the baseline SENet, with the same number of parameters and the same computational cost. Our code and models are publicly available at https://github.com/cfzd/FcaNet.

Dynamic inference networks, aimed at promoting computational efficiency, go along an adaptive executing path for a given sample. Prevalent methods typically assign a router for each convolutional block and sequentially make block-by-block executing decisions, without considering the relations during the dynamic inference. In this paper, we model the relations for dynamic inference from two aspects: the routers and the samples. We design a novel type of router called the relational router to model the relations among routers for a given sample. In principle, the current relational router aggregates the contextual features of preceding routers by graph convolution and propagates its router features to subsequent ones, making the executing decision for the current block in a long-range manner. Furthermore, we model the relation between samples by introducing a Sample Relation Module (SRM), encouraging correlated samples to go along correlated executing paths. As a whole, we call our method the Relational Dynamic Inference Network (RDI-Net). Extensive experiments on CIFAR-10/100 and ImageNet show that RDI-Net achieves state-of-the-art performance and computational cost reduction.

Spectral clustering has been widely used in various aspects, especially the machine learning fields. Clustering with similarity matrix and low-dimensional representation of data is the main reason of its promising performance shown in spectral clustering. However, such similarity matrix and low-dimensional representation directly derived from input data may not always hold when the data are high dimensional and has complex distribution. First, the similarity matrix simply based on the distance measurement might not be suitable for all kinds of data. Second, the low-dimensional representation might not be able to reflect the manifold structure of the original data. In this brief, we propose a novel linear space embedded clustering method, which uses adaptive neighbors to address the above-mentioned problems. Linearity regularization is used to make the data representation a linear embedded spectral. We also use adaptive neighbors to optimize the similarity matrix and clustering results simultaneously. Extensive experimental results show promising performance compared with the other state-of-the-art algorithms.

Representing high-volume and high-order data is an essential problem, especially in machine learning field. Although existing two-dimensional (2D) discriminant analysis achieves promising performance, the single and linear projection features make it difficult to analyze more complex data. In this paper, we propose a novel convolutional two-dimensional linear discriminant analysis (2D LDA) method for data representation. In order to deal with nonlinear data, a specially designed Convolutional Neural Networks (CNN) is presented, which can be proved having the equivalent objective function with common 2D LDA. In this way, the discriminant ability can benefit from not only the nonlinearity of Convolutional Neural Networks, but also the powerful learning process. Experiment results on several datasets show that the proposed method performs better than other state-of-the-art methods in terms of classification accuracy.