Abstract:Despite the recent advancements in 3D object detection, the conventional 3D point cloud object detection algorithms have been found to exhibit limited accuracy for the detection of small objects. To address the challenge of poor detection of small-scale objects, this paper adopts the PointPillars algorithm as the baseline model and proposes a two-stage 3D target detection approach. As a cutting-edge solution, point cloud processing is performed using Transformer models. Additionally, a redefined attention mechanism is introduced to further enhance the detection capabilities of the algorithm. In the first stage, the algorithm uses PointPillars as the baseline model. The central concept of this algorithm is to transform the point cloud space into equal-sized columns. During the feature extraction stage, when the features from all cylinders are transformed into pseudo-images, the proposed algorithm incorporates attention mechanisms adapted from the Squeeze-and-Excitation (SE) method to emphasize and suppress feature information. Furthermore, the 2D convolution of the traditional backbone network is replaced by dynamic convolution. Concurrently, the addition of the attention mechanism further improves the feature representation ability of the network. In the second phase, the candidate frames generated in the first phase are refined using a Transformer-based approach. The proposed algorithm applies channel weighting in the decoder to enhance channel information, leading to improved detection accuracy and reduced false detections. The encoder constructs the initial point features from the candidate frames for encoding. Meanwhile, the decoder applies channel weighting to enhance the channel information, thereby improving the detection accuracy and reducing false detections. In the KITTI dataset, the experimental results verify the effectiveness of this method in small objects detection. Experimental results show that the proposed method significantly improves the detection capability of small objects compared with the baseline PointPillars. In concrete terms, in the moderate difficulty detection category, cars, pedestrians, and cyclists average precision (AP) values increased by 5.30%, 8.1%, and 10.6%, respectively. Moreover, the proposed method surpasses existing mainstream approaches in the cyclist category.

HCT-Det: a Hybrid CNN-transformer Architecture for 3D Object Detection from Point Clouds

Improving 3D Object Detection with Channel-wise Transformer

PVT-SSD: Single-Stage 3D Object Detector with Point-Voxel Transformer

Anchor-Based Transformer for Temporal LiDAR 3D Object Detection

CT3D++: Improving 3D Object Detection with Keypoint-induced Channel-wise Transformer

Improved 3D Object Detection Based on PointPillars

An Efficient Point Cloud-based 3D Single Stage Object Detector*

MSIT-Det: Multi-Scale Feature Aggregation with Iterative Transformer Networks for 3D Object Detection

Transformer-Based Global PointPillars 3D Object Detection Method

Dynamic graph transformer for 3D object detection

CAT-Det: Contrastively Augmented Transformer for Multi-modal 3D Object Detection

TSSTDet: Transformation-Based 3-D Object Detection via a Spatial Shape Transformer

HybridPillars: Hybrid Point-Pillar Network for Real-time Two-stage 3D Object Detection

3D point cloud object detection algorithm based on Transformer

From Multi-View to Hollow-3D: Hallucinated Hollow-3D R-CNN for 3D Object Detection

Structure Aware Single-Stage 3D Object Detection From Point Cloud

DVST: Deformable Voxel Set Transformer for 3D Object Detection from Point Clouds

Context-Aware Dynamic Feature Extraction for 3D Object Detection in Point Clouds

FusionViT: Hierarchical 3D Object Detection via LiDAR-Camera Vision Transformer Fusion

Cross Modal Transformer: Towards Fast and Robust 3D Object Detection

Pillar-based multilayer pseudo-image 3D object detection