Although many deep-learning-based methods have achieved considerable detection performance for pedestrians with high visibility, their overall performances are still far from satisfactory, especially when heavily occluded instances are included. In this research, we have developed a novel pedestrian detector using a deformable attention-guided network (DAGN). Considering that pedestrians may be deformed with occlusions or under diverse poses, we have designed a deformable convolution with an attention module (DCAM) to sample from non-rigid locations, and obtained the attention feature map by aggregating global context information. Furthermore, the loss function was optimized to get accurate detection bounding boxes, by adopting complete-IoU loss for regression, and the distance IoU-NMS was used to refine the predicted boxes. Finally, a preprocessing technique based on tone mapping was applied to cope with the low visibility cases due to poor illumination. Extensive evaluations were conducted on three popular traffic datasets. Our method could decrease the log-average miss rate ($M{R}^{-2}$) by 12.44% and 7.8%, respectively, for the heavy occlusion and overall cases, when compared to the published state-of-the-art results of the Caltech pedestrian dataset. Of the CityPersons and EuroCity Persons datasets, our proposed method outperformed the current best results by about 5% in $M{R}^{-2}$ for the heavy occlusion cases.

Owing to its wide range of applications, 3D object detection has attracted increasing attention in computer vision tasks. Most existing 3D object detection methods are based on Lidar point cloud data. However, these methods have some limitations in localization consistency and classification confidence, due to the irregularity and sparsity of Light Detection and Ranging (LiDAR) point cloud data. Inspired by the complementary characteristics of Lidar and camera sensors, we propose a new end-to-end learnable framework named Point-Image Fusion Network (PIFNet) to integrate the LiDAR point cloud and camera images. To resolve the problem of inconsistency in the localization and classification, we designed an Encoder-Decoder Fusion (EDF) module to extract the image features effectively, while maintaining the fine-grained localization information of objects. Furthermore, a new effective fusion module is proposed to integrate the color and texture features from images and the depth information from the point cloud. This module can enhance the irregularity and sparsity problem of the point cloud features by capitalizing the fine-grained information from camera images. In PIFNet, each intermediate feature map is fed into the fusion module to be integrated with its corresponding point-wise features. Furthermore, point-wise features are used instead of voxel-wise features to reduce information loss. Extensive experiments using the KITTI dataset demonstrate the superiority of PIFNet over other state-of-the-art methods. Compared with several state-of-the-art methods, our approach outperformed by 1.97% in mean Average Precision (mAP) and by 2.86% in Average Precision (AP) for the hard cases on the KITTI 3D object detection benchmark.

Three-dimensional (3D) object detection is essential in autonomous driving. Three-dimensional (3D) Lidar sensor can capture three-dimensional objects, such as vehicles, cycles, pedestrians, and other objects on the road. Although Lidar can generate point clouds in 3D space, it still lacks the fine resolution of 2D information. Therefore, Lidar and camera fusion has gradually become a practical method for 3D object detection. Previous strategies focused on the extraction of voxel points and the fusion of feature maps. However, the biggest challenge is in extracting enough edge information to detect small objects. To solve this problem, we found that attention modules are beneficial in detecting small objects. In this work, we developed Frustum ConvNet and attention modules for the fusion of images from a camera and point clouds from a Lidar. Multilayer Perceptron (MLP) and tanh activation functions were used in the attention modules. Furthermore, the attention modules were designed on PointNet to perform multilayer edge detection for 3D object detection. Compared with a previous well-known method, Frustum ConvNet, our method achieved competitive results, with an improvement of 0.27%, 0.43%, and 0.36% in Average Precision (AP) for 3D object detection in easy, moderate, and hard cases, respectively, and an improvement of 0.21%, 0.27%, and 0.01% in AP for Bird’s Eye View (BEV) object detection in easy, moderate, and hard cases, respectively, on the KITTI detection benchmarks. Our method also obtained the best results in four cases in AP on the indoor SUN-RGBD dataset for 3D object detection.