Comparison of Autonomy and Study of Deep Learning Tools for Object Detection in Autonomous Self Driving Vehicles

Johari, Ayoush; Swami, Preety D.

doi:10.1109/idea49133.2020.9170659

Cited by 8 publications

(2 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Object detection is a computer vision task that involves both localizing and classifying objects within an image, where each detected object is wrapped by a bounding box. Object detection is widely used in several applications, such as self-driving technologies [ 38 ], surveillance and security [ 39 ] and robotics [ 40 ], among others.…”

Section: Related Workmentioning

confidence: 99%

Litter Detection with Deep Learning: A Comparative Study

Córdova

Pinto

Hellevik

et al. 2022

Sensors

View full text Add to dashboard Cite

Pollution in the form of litter in the natural environment is one of the great challenges of our times. Automated litter detection can help assess waste occurrences in the environment. Different machine learning solutions have been explored to develop litter detection tools, thereby supporting research, citizen science, and volunteer clean-up initiatives. However, to the best of our knowledge, no work has investigated the performance of state-of-the-art deep learning object detection approaches in the context of litter detection. In particular, no studies have focused on the assessment of those methods aiming their use in devices with low processing capabilities, e.g., mobile phones, typically employed in citizen science activities. In this paper, we fill this literature gap. We performed a comparative study involving state-of-the-art CNN architectures (e.g., Faster RCNN, Mask-RCNN, EfficientDet, RetinaNet and YOLO-v5), two litter image datasets and a smartphone. We also introduce a new dataset for litter detection, named PlastOPol, composed of 2418 images and 5300 annotations. The experimental results demonstrate that object detectors based on the YOLO family are promising for the construction of litter detection solutions, with superior performance in terms of detection accuracy, processing time, and memory footprint.

show abstract

Section: Related Workmentioning

confidence: 99%

Litter Detection with Deep Learning: A Comparative Study

Córdova

Pinto

Hellevik

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…For VD, most researchers focused on convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep reinforcement learning (DRL), which are the most commonly, used DL approaches in autonomous driving [17]. Convolutional neural network (CNN)-based approaches, in particular, have demonstrated promising results in detecting pedestrians, vehicles, and other objects in ADWC [18,19]. However, even with the use of CNNs for VD, real-time VD in a driving environment still needs to be improved.…”

Section: Introductionmentioning

confidence: 99%

An Optimal Feature Selection and Mrblstm-based Autonomous Vehicle Detection System in Challenging Weather Conditions

Vellaidurai,

Rathinam

2023

Preprint

View full text Add to dashboard Cite

Recently, self-driving cars with automation procedures such as queuing aid in traffic jams, parking assistance, lane-keep guidance, and crash evasion have been initiated. These automated driving systems and sophisticated video traffic surveillance methods laboriously rely on cameras and sensor fusion technologies. The poor performance of AVs and driver assistance systems in adverse weather conditions (ADWC), which might include snow, fog, rain, and hail, is one of the most serious problems in the development of these technologies. In order to detect items like vehicles and pedestrians in difficult situations like bad weather, it is vital to design the methods to do so. This paper proposes a deep learning (DL) model, namely a Modified ReLU-based Bidirectional Long Short Term Memory (MRBLSTM) for autonomous vehicle detection (VD) in ADWC. It is mainly composed of six stages. Initially, the proposed system collects the data from the publicly available dataset for VD. After that, image restoration is done as a preprocessing step on the collected data using the Retinex algorithm. Next, the Modified Canny Edge Detection Operator (MCO) algorithm preserves the edges from the restored images with the help of a bilateral filter and Otsu’s approach. Then, using feature extraction models for HoG, HaaR-like features, LBP, and SIFT, the features are retrieved from the edge-detected images. The optimal features are selected from the extracted features to lower the dimensions of the features using the Cauchy mutation-included Coyotes Optimization Algorithm (CMCOA). Finally, the vehicles are detected using the MRBLSTM. The proposed strategy has experimented on the DAWN dataset, and testing outcomes have demonstrated the proposed technique's efficacy, which surpasses state-of-the-art VD systems under ADWC.

show abstract