Fast-BEV: A Fast and Strong Bird's-Eye View Perception Baseline.

Recently, perception task based on Bird's-Eye View (BEV) representation has drawn more and more attention, and BEV representation is promising as the foundation for next-generation Autonomous Vehicle (AV) perception. However, most existing BEV solutions either require considerable resources to execute on-vehicle inference or suffer from modest performance. This paper proposes a simple yet effective framework, termed Fast-BEV, which is capable of performing faster BEV perception on the on-vehicle chips. Towards this goal, we first empirically find that the BEV representation can be sufficiently powerful without expensive transformer based transformation nor depth representation. Our Fast-BEV consists of five parts, We innovatively propose (1) a lightweight deploymentfriendly view transformation which fast transfers 2D image feature to 3D voxel space, (2) an multi-scale image encoder which leverages multi-scale information for better performance, (3) an efficient BEV encoder which is particularly designed to speed up on-vehicle inference. We further introduce (4) a strong data augmentation strategy for both image and BEV space to avoid over-fitting, (5) a multiframe feature fusion mechanism to leverage the temporal information. Among them, (1) and (3) enable Fast-BEV to be fast inference and deployment friendly on the on-vehicle chips, (2), (4) and (5) ensure that Fast-BEV has competitive performance. All these make Fast-BEV a solution with high performance, fast inference speed, and deployment-friendly on the on-vehicle chips of autonomous driving. Through experiments, on 2080Ti platform, our R50 model can run 52.6 FPS with 47.3% NDS on the nuScenes validation set, exceeding the 41.3 FPS and 47.5% NDS of the BEVDepth-R50 model [1] and 30.2 FPS and 45.7% NDS of the BEVDet4D-R50 model [2]. Our largest model (R101@900x1600) establishes a competitive 53.5% NDS on the nuScenes validation set. We further develop a benchmark with considerable accuracy and efficiency on current popular on-vehicle chips. The code is released at: https://github.com/Sense-GVT/FastBEV.

Centennial Records of Microplastics in Lake Cores in Huguangyan Maar Lake, China.

Microplastic records from lake cores can reconstruct the plastic pollution history. However, the associations between anthropogenic activities and microplastic accumulation are not well understood. Huguangyan Maar Lake (HML) is a deep-enclosed lake without inlets and outlets, where the sedimentary environment is ideal for preserving a stable and historical microplastic record. Microplastic (size: 10-500 µm) characteristics in the HML core were identified using the Laser Direct Infrared Imaging system. The earliest detectable microplastics appeared unit in 1955 (1.1 items g-1). The microplastic abundance ranged from n.d. to 615.2 items g-1 in 1955-2019 with an average of 134.9 items g-1. The abundance declined slightly during the 1970s and then increased rapidly after China's Reform and Opening Up in 1978. Sixteen polymer types were detectable, with polyethylene and polypropylene dominating, accounting for 23.5 and 23.3% of the total abundance, and the size at 10-100 µm accounted for 80%. Socioeconomic factors dominated the microplastic accumulation based on the random forest modeling, and the contributions of GDP per capita, plastic-related industry yield, and total crop yield were, respectively, 13.9, 35.1, and 9.3% between 1955-2019. The total crop yield contribution further increased by 1.7% after 1978. Coarse sediment particles increased with soil erosion exacerbated microplastics discharging into the sediment.

Delving Into the Devils of Bird's-Eye-View Perception: A Review, Evaluation and Recipe.

Learning powerful representations in bird's-eye-view (BEV) for perception tasks is trending and drawing extensive attention both from industry and academia. Conventional approaches for most autonomous driving algorithms perform detection, segmentation, tracking, etc., in a front or perspective view. As sensor configurations get more complex, integrating multi-source information from different sensors and representing features in a unified view come of vital importance. BEV perception inherits several advantages, as representing surrounding scenes in BEV is intuitive and fusion-friendly; and representing objects in BEV is most desirable for subsequent modules as in planning and/or control. The core problems for BEV perception lie in (a) how to reconstruct the lost 3D information via view transformation from perspective view to BEV; (b) how to acquire ground truth annotations in BEV grid; (c) how to formulate the pipeline to incorporate features from different sources and views; and (d) how to adapt and generalize algorithms as sensor configurations vary across different scenarios. In this survey, we review the most recent works on BEV perception and provide an in-depth analysis of different solutions. Moreover, several systematic designs of BEV approach from the industry are depicted as well. Furthermore, we introduce a full suite of practical guidebook to improve the performance of BEV perception tasks, including camera, LiDAR and fusion inputs. At last, we point out the future research directions in this area. We hope this report will shed some light on the community and encourage more research effort on BEV perception.

CycleMLP: A MLP-Like Architecture for Dense Visual Predictions.

This article presents a simple yet effective multilayer perceptron (MLP) architecture, namely CycleMLP, which is a versatile neural backbone network capable of solving various tasks of dense visual predictions such as object detection, segmentation, and human pose estimation. Compared to recent advanced MLP architectures such as MLP-Mixer (Tolstikhin et al. 2021), ResMLP (Touvron et al. 2021), and gMLP (Liu et al. 2021), whose architectures are sensitive to image size and are infeasible in dense prediction tasks, CycleMLP has two appealing advantages: 1) CycleMLP can cope with various spatial sizes of images; 2) CycleMLP achieves linear computational complexity with respect to the image size by using local windows. In contrast, previous MLPs have O(N2) computational complexity due to their full connections in space. 3) The relationship between convolution, multi-head self-attention in Transformer, and CycleMLP are discussed through an intuitive theoretical analysis. We build a family of models that can surpass state-of-the-art MLP and Transformer models e.g., Swin Transformer (Liu et al. 2021), while using fewer parameters and FLOPs. CycleMLP expands the MLP-like models' applicability, making them versatile backbone networks that achieve competitive results on dense prediction tasks For example, CycleMLP-Tiny outperforms Swin-Tiny by 1.3% mIoU on ADE20 K dataset with fewer FLOPs. Moreover, CycleMLP also shows excellent zero-shot robustness on ImageNet-C dataset.

Three-dimensional spatial prediction of Zn in the soil of a former tire manufacturing plant using machine learning and readily attainable multisource auxiliary data.

Pollutants in the soil of industrial site are often highly heterogeneously distributed, which brought a challenge to accurately predict their three-dimensional (3D) spatial distributions. Here we attempt to create effective 3D prediction models using machine learning (ML) and readily attainable multisource auxiliary data for improving the prediction accuracy of highly heterogeneous Zn in the soil of a small-size industrial site. Using raw covariates from functional area layout, stratigraphic succession, and electrical resistivity tomography, and derived covariates of the raw covariates as predictors, we created 6 individual and 2 ensemble models for Zn, based on ML algorithms such as k-nearest neighbors, random forest, and extreme gradient boosting, and the stacking approach in ensemble ML. Results showed that the overall 3D spatial patterns of Zn predicted by individual and ensemble ML models, inverse distance weighting (IDW), and ordinary Kriging (OK) were similar, but their predictive performances differed significantly. The ensemble model with raw and derived covariates had the highest accuracy in representing the complex 3D spatial patterns of Zn (R2 = 0.45, RMSE = 344.80 mg kg-1), compared to the accuracies of individual ML models (R2 = 0.27-0.44, RMSE = 396.75-348.56 mg kg-1), OK (R2 = 0.33, RMSE = 381.12 mg kg-1), and IDW interpolation (R2 = 0.25, RMSE = 402.94 mg kg-1). Besides, the prediction accuracy gains of incorporating derived covariates were higher than adopting ensemble ML instead of single ML algorithm. These results highlighted the importance of developing derived covariates whilst adopting ML in predicting the 3D distribution of highly heterogeneous pollutant in the soil of small-size industrial site.

Deeply Unsupervised Patch Re-Identification for Pre-training Object Detectors.

Unsupervised pre-training aims at learning transferable features that are beneficial for downstream tasks. However, most state-of-the-art unsupervised methods concentrate on learning global representations for image-level classification tasks instead of discriminative local region representations, which limits their transferability to region-level downstream tasks, such as object detection. To improve the transferability of pre-trained features to object detection, we present Deeply Unsupervised Patch Re-ID (DUPR), a simple yet effective method for unsupervised visual representation learning. The patch Re-ID task treats individual patch as a pseudo-identity and contrastively learns its correspondence in two views, enabling us to obtain discriminative local features for object detection. Then the proposed patch Re-ID is performed in a deeply unsupervised manner, appealing to object detection, which usually requires multi-level feature maps. Extensive experiments demonstrate that DUPR outperforms state-of-the-art unsupervised pre-trainings and even the ImageNet supervised pre-training on various downstream tasks related to object detection.

PAN++: Towards Efficient and Accurate End-to-End Spotting of Arbitrarily-Shaped Text.

Scene text detection and recognition have been well explored in the past few years. Despite the progress, efficient and accurate end-to-end spotting of arbitrarily-shaped text remains challenging. In this work, we propose an end-to-end text spotting framework, termed PAN++, which can efficiently detect and recognize text of arbitrary shapes in natural scenes. PAN++ is based on the kernel representation that reformulates a text line as a text kernel (central region) surrounded by peripheral pixels. By systematically comparing with existing scene text representations, we show that our kernel representation can not only describe arbitrarily-shaped text but also well distinguish adjacent text. Moreover, as a pixel-based representation, the kernel representation can be predicted by a single fully convolutional network, which is very friendly to real-time applications. Taking the advantages of the kernel representation, we design a series of components as follows: 1) a computationally efficient feature enhancement network composed of stacked Feature Pyramid Enhancement Modules (FPEMs); 2) a lightweight detection head cooperating with Pixel Aggregation (PA); and 3) an efficient attention-based recognition head with Masked RoI. Benefiting from the kernel representation and the tailored components, our method achieves high inference speed while maintaining competitive accuracy. Extensive experiments show the superiority of our method. For example, the proposed PAN++ achieves an end-to-end text spotting F-measure of 64.9 at 29.2 FPS on the Total-Text dataset, which significantly outperforms the previous best method. Code will be available at: git.io/PAN.

Spatiotemporal variation and sources of soil heavy metals along the lower reaches of Yangtze River, China.

Excessive accumulation of soil heavy metals (HMs) result in the deterioration of soil quality and reduction of agricultural productivity and safety. The accumulation status, temporal change, and sources of soil HMs were determined by large-scale field surveys in 2014 and 2019 in rapid urbanization and industrialization area along the lower reaches of the Yangtze River, China. Eighty-two surface soil samples were collected in 2014 and ninety-five surface soil samples and seven soil profiles (0-100 cm) were collected in 2019. The mean concentrations (in, mg kg-1) of As (10.17), Cd (0.33), Cr (86.38), Cu (38.22), Hg (0.11), Ni (37.67), Pb (43.95), and Zn (113.15) were greater than the corresponding background values. The concentrations of these 8 HMs significantly varied with site-specific distributions depending on nearby landscape patterns with decreasing order: agricultural soil around industrial > agricultural soil > fallow soil. Cd and Hg were found to be priority pollutants due to their greater accumulations in this study area. Combined analyses of principal component analysis and positive matrix factorization model addressed source apportionment of soil HMs. Industrial activities, parent materials, and agricultural and traffic activities were three major sources and their contributions were 35.56%, 35.20%, and 29.23%, respectively. The concentrations of soil As, Cd, Cr and Pb increased with time. This study elucidates how changes in land uses and time affect soil HMs and provides reasonable suggestions for the effective reduction of HM contamination in economically and industrially developed areas of China, and elsewhere.

PolarMask++: Enhanced Polar Representation for Single-Shot Instance Segmentation and Beyond.

Reducing complexity of the pipeline of instance segmentation is crucial for real-world applications. This work addresses this problem by introducing an anchor-box free and single-shot instance segmentation framework, termed PolarMask++, which reformulates the instance segmentation problem as predicting the contours of objects in the polar coordinate, leading to several appealing benefits. (1) The polar representation unifies instance segmentation (masks) and object detection (bounding boxes) into a single framework, reducing the design and computational complexity. (2) We carefully design two modules (soft polar centerness and polar IoU loss) to sample high-quality center examples and optimize polar contour regression, making the performance of PolarMask++ does not depend on the bounding box prediction and thus more efficient in training. (3) PolarMask++ is fully convolutional and can be easily embedded into most off-the-shelf detectors. To further improve the accuracy of the framework, a Refined Feature Pyramid is introduced to improve the feature representation at different scales. Extensive experiments demonstrate the effectiveness of PolarMask++, which achieves competitive results on COCO dataset, and new state-of-the-art results on text detection and cell segmentation datasets. We hope polar representation can provide a new perspective for designing algorithms to solve single-shot instance segmentation. Code is released at: github.com/xieenze/PolarMask.

Spatio-Temporal Change and Pollution Risk of Agricultural Soil Cadmium in a Rapidly Industrializing Area in the Yangtze Delta Region of China.

The impacts of rapid industrialization on agricultural soil cadmium (Cd) accumulation and their potential risks have drawn major attention from the scientific community and decision-makers, due to the high toxicity of Cd to animals and humans. A total of 812 topsoil samples (0⁻20 cm) was collected from the southern parts of Jiangsu Province, China, in 2000 and 2015, respectively. Geostatistical ordinary kriging and conditional sequential Gaussian simulation were used to quantify the changes in spatial distributions and the potential risk of Cd pollution between the two sampling dates. Results showed that across the study area, the mean Cd concentrations increased from 0.110 mg/kg in 2000 to 0.196 mg/kg in 2015, representing an annual average increase of 5.73 µg/kg. Given a confidence level of 95%, areas with significantly-increased Cd covered approximately 12% of the study area. Areas with a potential risk of Cd pollution in 2000 only covered 0.009% of the study area, while this figure increased to 0.75% in 2015. In addition, the locally concentrating trend of soil Cd pollution risk was evident after 15 years. Although multiple factors contributed to this elevated Cd pollution risk, the primary reason can be attributed to the enhanced atmospheric deposition and industrial waste discharge resulting from rapid industrialization, and the quick increase of traffic and transportation associated with rapid urbanization.

Soil mercury accumulation, spatial distribution and its source identification in an industrial area of the Yangtze Delta, China.

Understanding soil mercury (Hg) accumulation, spatial distribution, and its sources is crucial for effective regulation of Hg emissions. We chose a study area covering approximately 100 km2 representing one of the rapid growing industrial towns of the Yangtze River Delta (YRD), China, to explore soil Hg accumulation. In surface soil, total Hg ranged from 310 to 3760 µg/kg, and 53% samples exceeded the most generous Chinese soil critical value (1500 µg/kg). Hg concentration in rice ranged from 10 to 40 µg/kg, and 43% samples exceeded the regulatory critical value (20 µg/kg). Total Hg concentrations in soil profiles gradually decreased, reaching background levels up to 60 cm profile depth. Meanwhile, proportions of mobile, semi-mobile and non-mobile Hg to total Hg at every soil depth were similar, leading us to deduce that soil Hg has accumulated in this area over a long period. Total and bioavailable Hg in topsoil exhibited the highest concentrations in the center of the study area, and radially decreased towards the periphery, which might be explained by the distribution of industry and the prevailing wind. To trace the Hg sources, we selected soil and atmospheric dust samples for isotope analysis. Hg isotopic composition of surface soil (δ202Hg = -0.29 ±â€¯0.10‰ and Δ199Hg = 0.03 ±â€¯0.03‰) was close to that of atmospheric dust (δ202Hg = -0.54 ±â€¯0.10‰ and Δ199Hg = 0.03 ±â€¯0.05‰), but considerably different from Hg isotopic composition in subsoil (δ202Hg = -0.90 ±â€¯0.09‰ and Δ199Hg = -0.04 ±â€¯0.04‰). Thus, we speculated that atmospheric deposition could change Hg isotopic composition in topsoil. Our findings suggest that when Hg atmospheric dust deposition changes Hg levels in surface soil, soil remediation, and crop safety might be compromised.