PLOT: a 3D point cloud object detection network for autonomous driving

In this paper, we present a hardware-software implementation of a deep neural network for object detection based on a point cloud obtained by a LiDAR sensor. The Brevitas / PyTorch tools were used for network quantisation and the FINN tool for hardware implementation in the reprogrammable Zynq UltraScale+ MPSoC device. The PointPillars network was used in the research, as it is a reasonable compromise between detection accuracy and calculation complexity. The obtained results show that quite a significant computation precision limitation along with a few network architecture simplifications allows the solution to be implemented on an heterogeneous embedded platform with reasonable detection accuracy.

Traditionally, point cloud-based 3D object detectors are trained on annotated, non-sequential samples taken from driving sequences (e.g. the KITTI dataset). However, by doing this, the developed algorithms renounce to exploit any dynamic information from the driving sequences. It is reasonable to think that this information, which is available at test time when deploying the models in the experimental vehicles, could have significant predictive potential for the object detection task. To study the advantages that this kind of information could provide, we construct a dataset of dynamic occupancy grid maps from the raw KITTI dataset and find the correspondence to each of the KITTI 3D object detection dataset samples. By training a Lidar-based state-of-the-art 3D object detector with and without the dynamic information we get insights into the predictive value of the dynamics. Our results show that having access to the environment dynamics improves by 27% the ability of the detection algorithm to predict the orientation of smaller obstacles such as pedestrians. Furthermore, the 3D and bird's eye view bounding box predictions for pedestrians in challenging cases also see a 7% improvement. Qualitatively speaking, the dynamics help with the detection of partially occluded and far-away obstacles. We illustrate this fact with numerous qualitative prediction results.

Object detection in point clouds is a critical component in most autonomous driving systems. In this paper, in order to improve the effectiveness of image feature extraction and the accuracy of detection of point clouds, a pillar-based 3D point cloud object detection algorithm with multiattention mechanism is proposed, which includes three attention mechanisms SOCA, SOPA, and SAPI. The results show that the recognition accuracy of the optimized algorithm for cars, pedestrians, and cyclists on KITTI dataset is significantly improved on the detection benchmarks of BEV and 3D. Despite using only LiDAR, our algorithm outperforms PointPillars, which is one of the state-of-the-art algorithms for 3D object detection, with respect to both 3D and BEV view KITTI benchmarks while maintaining a relatively competitive speed.

Light Detection And Ranging (LiDAR) sensors can generate a number of sequential 3D point clouds, which are widely deployed in many real-world systems. 3D object detection in point clouds, is one of the most fundamental tasks. Unfortunately, the existing 3D object detection methods degrade in snowy weather, because in that situation the annotated samples are difficult to collect. To solve this issue, we propose a novel GAN-based Snowfall Point-cloud AugmentOR (GAN spaor ) to generate high-quality synthetic snowfall point clouds as augmentations. The basic idea of GAN spaor is to transfer annotated point clouds to snowfall versions by simultaneously learning the global style of real snowfall point clouds and the local details of physics-induced ones. Our framework fuses data-driven and physical modeling methods for rapidly generating data in snowy weather. To evaluate the effectiveness of GAN spaor , we employ a number of recent 3D object detection methods and train them by using the synthetic samples of GAN spaor as auxiliary augmentations. Moreover, we conduct a comparative analysis of the characteristics of the data distributions of the snowy point clouds synthesized by GAN spaor . Experimental results demonstrate that GAN spaor can improve the performance of 3D object detection methods compared with other existing snowfall point cloud simulators.

In the recent 3D object detection methods for point clouds, the combination of point-based methods and voxel-based methods is gradually becoming a trend. Point-based methods retain the accurate position and pose information in the raw points and voxel-based methods get multi-scale structure information through the 3D backbone. However, because of the sparsity and irregularity of point clouds, both representations ignore the context information, which is important for the detection of sparse and small objects. To solve this problem, we propose a multi-stream feature aggregation network to extract features from three representations of the point cloud for object detection. Specifically, we exploit multi-stream features extracted from point, voxel, and perspective view (PV) respectively on a parallel way, where the complementary information between different perspectives can be used to enrich the feature representations, especially for the perspective view containing rich semantic context information. Secondly, to eliminate redundant information and better exploit the correlation between different feature representations, we design an attention-based multi-stream feature fusion module (MSFF) to combine features from three information streams. Besides, we introduce a new voxel RoI pooling with the self-attention in the second refinement stage, which can further strengthen the connection between local features in the proposal to obtain accurate classification and localization predictions. Our method achieves progressive results on the KITTI dataset, especially in the cyclist category, which improves the baseline significantly by 5.56%, 4.73%, 5.16% AP in the test set for easy, moderate, and hard levels respectively. Code will be available at https://github.com/june2678/MR F.

3D object detection from raw and sparse point clouds has been far less treated to date, compared with its 2D counterpart. In this paper, we propose a novel framework called FVNet for 3D front-view proposal generation and object detection from point clouds. It consists of two stages: generation of front-view proposals and estimation of 3D bounding box parameters. Instead of generating proposals from camera images or bird's-eye-view maps, we first project point clouds onto a cylindrical surface to generate front-view feature maps which retains rich information. We then introduce a proposal generation network to predict 3D region proposals from the generated maps and further extrude objects of interest from the whole point cloud. Finally, we present another network to extract the point-wise features from the extruded object points and regress the final 3D bounding box parameters in the canonical coordinates. Our framework achieves real-time performance with 12ms per point cloud sample. Extensive experiments on the 3D detection benchmark KITTI show that the proposed architecture outperforms state-of-the-art techniques which take either camera images or point clouds as input, in terms of accuracy and inference time.

Accurate object detection is a fundamental requirement for autonomous systems to operate in dynamic urban environments. Considering the intricacy of the environment and the occlusion of objects, various point cloud based three-dimensional (3D) object detection methods have been proposed, such as point-based or voxel-based methods. In this paper, a feature selection mechanism is proposed in a voxel-based method to generate bird’s eye view (BEV) feature maps from the original point cloud. For the 3D Backbone, the U-shaped structure and single scale feature selection module are combined. After combining high-level semantics and low-level fine-grained features, the optimized features after the BEV feature maps are applied to region of interest (RoI) refinement, so that the voxel features can better serve the subsequent 3D object detection. The experimental results on KITTI dataset show higher 3D object detection accuracy compared to the state-of-the-art 3D detection methods, which reflects the effectiveness of the proposed architecture.

Light detection and ranging (LiDAR) is widely used in the automotive industry as it can provide point cloud information containing precise distances. Three-dimensional (3D) object detection based on LiDAR point clouds is significant for environment perception tasks. However, feature learning for point clouds remains challenging. This paper proposes a two-stage voxel-based LiDAR 3D object detector, referred to as density-aware and neighborhood attention (DenNet), that focuses on the neighborhood information of objects. DenNet mainly integrates two modules: voxel density-aware (VDA) and neighborhood attention (NA). VDA introduces density information of the point cloud. Here, point cloud density information was added as voxel features in the voxel-based framework to alleviate the information loss during voxelization. Additionally, to extract neighbor information, the characteristics of 3D objects were analyzed for traffic scenes. The NA mechanism was adopted, which localizes the receptive field for each query to its nearest neighboring points. DenNet yielded competitive results, as compared with state-of-the-art methods, for the KITTI and One Million Scenes (ONCE) datasets; notably, it afforded an improvement of 3.96% relative to the baseline mean average precision on the more challenging ONCE dataset.

Dense frustum-aware fusion for 3D object detection in perception systems

Dynamic graph transformer for 3D object detection

Segmenting or detecting objects in sparse Lidar point clouds are two important tasks in autonomous driving to allow a vehicle to act safely in its 3D environment. The best performing methods in 3D semantic segmentation or object detection rely on a large amount of annotated data. Yet annotating 3D Lidar data for these tasks is tedious and costly. In this context, we propose a self-supervised pretraining method for 3D perception models that is tailored to autonomous driving data. Specifically, we leverage the availability of synchronized and calibrated image and Lidar sensors in autonomous driving setups for distilling self-supervised pre-trained image representations into 3D models. Hence, our method does not require any point cloud nor image annotations. The keyingredient of our method is the use of superpixels which are used to pool 3D point features and 2D pixel features in visually similar regions. We then train a 3D network on the self-supervised task of matching these pooled point features with the corresponding pooled image pixel features. The advantages of contrasting regions obtained by superpixels are that: (1) grouping together pixels and points of visually coherent regions leads to a more meaningful contrastive task that produces features well adapted to 3D semantic segmentation and 3D object detection; (2) all the different regions have the same weight in the contrastive loss regardless of the number of 3D points sampled in these regions; (3) it mitigates the noise produced by incorrect matching of points and pixels due to occlusions between the different sensors. Extensive experiments on autonomous driving datasets demonstrate the ability of our image-to-Lidar distillation strategy to produce 3D representations that transfer well on semantic segmentation and object detection tasks.

Currently, there are many kinds of voxel-based multisensor 3D object detectors, while point-based multisensor 3D object detectors have not been fully studied. In this paper, we propose a new 3D two-stage object detection method based on point cloud and image fusion to improve the detection accuracy. To address the problem of insufficient semantic information of point cloud, we perform multiscale deep fusion of LiDAR point and camera image in a point-wise manner to enhance point features. Due to the imbalance of LiDAR points, the object point cloud in the long-distance area is sparse. We design a point cloud completion module to predict the spatial shape of objects in the candidate boxes and extract the structural information to improve the feature representation ability to further refine the boxes. The framework is evaluated on widely used KITTI and SUN-RGBD dataset. Experimental results show that our method outperforms all state-of-the-art point-based 3D object detection methods and has comparable performance to voxel-based methods as well.

Voxel is one of the common structural representation of 3D point cloud. Due to the sparsity of point cloud generated by light detection and ranging (LiDAR), there is the extreme imbalance in the foreground and background voxels. It decreases the accuracy of 3D object detection, has the negative effect on intelligent driving safety. To overcome this problem, we present a saliency prediction based 3D object detector SP-Det in this article. Although foreground voxels have the sufficient feature of object, it is difficult to localize the foreground region from voxel space with the larger background region. We design an auxiliary learning task, saliency prediction (SP). It benefits 3D detector in identifying the foreground region. SP task uses label diffusion to alleviate the label imbalance. It reduces the learning difficulty of saliency in voxel and bird's eye view (BEV) spaces. After that, to strengthen feature interaction from the sparse foreground region, we design saliency fusion (SF) module to fuse the learning result in SP task. It utilizes voxel and BEV saliency maps as progressive attention to resist the redundant feature from background region. To aggregate more foreground feature inside 3D and BEV region of interest (RoI), we design hybrid grid maps based RoI pooling (Hybrid-RoI pooling). Experiments are conducted in STF dataset. The adverse weather enlarges the sparsity of LiDAR point cloud, increasing the difficulty of object detection. SP-Det identifies and leverages foreground region, and achieves the performance better than the current methods. Hence, we believe that SP-Det benefits to LiDAR based 3D scene understanding in the adverse weather.

Multi-scale object detection in natural scenes is still challenging. To enhance the multi-scale perception capability, some algorithms combine the lower-level and higher-level information via multi-scale feature fusion strategies. However, the inherent spatial properties among instances and relations between foreground and background are ignored. In addition, the human-defined “center-based” regression quality evaluation strategy, predicting a high-to-low score based on a linear relationship with the distance to the center of ground-truth box, is not robust to scale-variant objects. In this work, we propose a Depth-Guided Progressive Network (DGPNet) for multi-scale object detection. Specifically, besides the prediction of classification and localization, the depth is estimated and used to guide the image features in a weighted manner to obtain a better spatial representation. Therefore, depth estimation and 2D object detection are simultaneously learned via a unified network, where the depth features are merged as auxiliary information into the detection branch to enhance the discrimination among multi-scale objects. Moreover, to overcome the difficulty of empirically fitting the localization quality function, high-quality predicted boxes on scale-variant objects are more adaptively obtained by an IoU-aware progressive sampling strategy. We divide the sampling process into two stages, i.e., “statistical-aware” and “IoU-aware”. The former selects thresholds for positive samples based on statistical characteristics of multi-scale instances, and the latter further selects high-quality samples by IoU on the basis of the former. Therefore, the final ranking scores better reflect the quality of localization. Experiments verify that our method outperforms state-of-the-art methods on the KINS and Cityscapes dataset.

Current point cloud object detection methods rely on expensive manual annotation. Utilizing contrastive learning for self-supervised pre-training on unlabeled large-scale point clouds can reduce annotation costs and improve model performance. However, selecting effective features for instance discrimination is crucial for contrastive learning. Previous methods have constructed instances for pre-training at different levels, such as points, proposals, and scenes, but the features of these instances differ from the objects to be detected. Considering that instance discrimination tasks based on object-level features align with downstream object detection tasks, we propose a novel and efficient self-supervised point cloud object detection pre-training framework called ObjectContrast. To learn more effective point cloud representations, this framework constructs two self-supervised pre-training modules: object-level instance discrimination contrast ObCo and bounding box geometric contrast prediction BoxCo . ObCo drives the model to learn general object representations to locate object foregrounds and determine categories. BoxCo enhances the model's geometric perception capabilities regarding the dimension and orientation of 3D bounding boxes. Extensive experiments on various detectors and datasets validate the efficiency and transferability of ObjectContrast. Compared with the state-of-the-art self-supervised pre-training methods, ObjectContrast demonstrates superior performance.