Image De-Raining for Driver Assistance Systems using U-Net based GAN

Ramwala, Ojas A.; Paunwala, Chirag N.; Paunwala, Mita

doi:10.1109/spicscon48833.2019.9065063

Cited by 6 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Current driver-assistance systems are gradually being equipped with more advanced technology. Most systems aim to provide parking assistance, forward collision warnings, lane-departure warnings, adaptive cruise control, and driver drowsiness detection [46].…”

Section: Advanced Driver-assistance Systemmentioning

confidence: 99%

Review of Autonomous Intelligent Vehicles for Urban Driving and Parking

Chan

Chin

2021

Electronics

View full text Add to dashboard Cite

With the concept of Internet-of-Things, autonomous vehicles can provide higher driving efficiency, traffic safety, and freedom for the driver to perform other tasks. This paper first covers enabling technology involving a vehicle moving out of parking, traveling on the road, and parking at the destination. The development of autonomous vehicles relies on the data collected for deployment in actual road conditions. Research gaps and recommendations for autonomous intelligent vehicles are included. For example, a sudden obstacle while the autonomous vehicle executes the parking trajectory on the road is discussed. Several aspects of social problems, such as the liability of an accident affecting the autonomous vehicle, are described. A smart device to detect abnormal driving behaviors to prevent possible accidents is briefly discussed.

show abstract

Section: Advanced Driver-assistance Systemmentioning

confidence: 99%

Review of Autonomous Intelligent Vehicles for Urban Driving and Parking

Chan

Chin

2021

Electronics

View full text Add to dashboard Cite

show abstract

“…As the training of deep CNNs suffer from vanishing gradient problem. Therefore, many previous works [3,4,25] used skip-connections in the generator to pass the gradient easily to prior layers of the encoder. Unfortunately, these skip-connections directly carry unwanted information from the inputs to the resultant images, hence affecting the visual quality of the constructing images.…”

Section: Related Workmentioning

confidence: 99%

“…These tasks have multiple applications in computer graphics, image processing, and computer vision. Image processing applications include: (i) image in-painting, where damaged parts of an image are restored [ 1 , 2 ], (ii) image de-raining where rain-streaks are removed from an input image to get rain-free image [ 3 , 4 ], (iii) image super-resolution where high-quality images are generated from similar degraded images [ 5 , 6 , 7 , 8 , 9 , 10 ]. Additional applications exist, however they are not constrained to image denoising [ 11 , 12 , 13 ], style transfer [ 14 ], image segmentation [ 15 ] and image colorization [ 16 , 17 ].…”

Section: Introductionmentioning

confidence: 99%

An Input-Perceptual Reconstruction Adversarial Network for Paired Image-to-Image Conversion

Khan

Jin

Ahmad

et al. 2020

Sensors

View full text Add to dashboard Cite

Image-to-image conversion based on deep learning techniques is a topic of interest in the fields of robotics and computer vision. A series of typical tasks, such as applying semantic labels to building photos, edges to photos, and raining to de-raining, can be seen as paired image-to-image conversion problems. In such problems, the image generation network learns from the information in the form of input images. The input images and the corresponding targeted images must share the same basic structure to perfectly generate target-oriented output images. However, the shared basic structure between paired images is not as ideal as assumed, which can significantly affect the output of the generating model. Therefore, we propose a novel Input-Perceptual and Reconstruction Adversarial Network (IP-RAN) as an all-purpose framework for imperfect paired image-to-image conversion problems. We demonstrate, through the experimental results, that our IP-RAN method significantly outperforms the current state-of-the-art techniques.

show abstract