A Nighttime Vehicle Detection Method with Attentive GAN for Accurate Classification and Regression

Abstract

Vehicle detection plays a vital role in the design of Automatic Driving System (ADS), which has achieved remarkable improvements in recent years. However, vehicle detection in night scenes still has considerable challenges for the reason that the vehicle features are not obvious and are easily affected by complex road lighting or lights from vehicles. In this paper, a high-accuracy vehicle detection algorithm is proposed to detect vehicles in night scenes. Firstly, an improved Generative Adversarial Network (GAN), named Attentive GAN, is used to enhance the vehicle features of nighttime images. Then, with the purpose of achieving a higher detection accuracy, a multiple local regression is employed in the regression branch, which predicts multiple bounding box offsets. An improved Region of Interest (RoI) pooling method is used to get distinguishing features in a classification branch based on Faster Region-based Convolutional Neural Network (R-CNN). Cross entropy loss is introduced to improve the accuracy of classification branch. The proposed method is examined with the proposed dataset, which is composed of the selected nighttime images from BDD-100k dataset (Berkeley Diverse Driving Database, including 100,000 images). Compared with a series of state-of-the-art detectors, the experiments demonstrate that the proposed algorithm can effectively contribute to vehicle detection accuracy in nighttime.

Apple-Net: A Model Based on Improved YOLOv5 to Detect the Apple Leaf Diseases

“…We used the attentive GAN algorithm [ 42 ] to reduce the raindrop noise of apple leaves. Attentive GAN is a raindrop removal method based on a single image.…”

GHE-Ensemble: Enhanced Hybrid Image Enhancement Model for Night Vision Multi-Object detection in Autonomous Vehicle

“…Many information like vehicle, road, traffic light, person will be lost due to poor lighting. 3 No accurate object is detected due to motion blurring, blurred images or partial occlusion [5]. 4 Night images does not receive the sufficient light due to which there will be more noise resulting in the failure of the multi object detection.…”

“…This method resulted in the 57.59% of LAMR (low average miss rate) with fusion and without fusion in the feature extraction. [5] Used BDD open-source dataset with generative adversial network for the enhance feature extraction on the road. In the deep learning framework used ROI pooling method which resulted in 45% average precision in the confidence level 75.…”

HMD-Net: A Vehicle Hazmat Marker Detection Benchmark

“…Therefore, reducing the accident rate by detecting and monitoring vehicles for hazmat transportation is a common concern for the government, manufacturers and carriers. In fact, there are many ways for vehicle monitoring [2,3]. In terms of vehicles for hazmat transportation, as required by the Standardization Administration of China, they must be installed and attached with corresponding warning markers, including triangle light boxes and license plates, as shown in Figure 1a and Figure 1b, respectively.…”