Nighttime Data Augmentation Using GAN for Improving Blind-Spot Detection

Lee, Hongjun; Ra, Moonsoo; Kim, Whoi-Yul

doi:10.1109/access.2020.2979239

Cited by 31 publications

(11 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the experimental results, DG-Net achieved gains of 8.3% and 10.3% mAP on Market-1501 and DukeMTMC-reID, respectively, indicating the advantage of the proposed joint learning. In [52], Lee et al proposed an any-time-of day camera-based BSD system and generated a synthetic nighttime side-rectilinear images to improve the nighttime performance of the hand-crafted feature-based BSD system. The framework with a conditional GAN for data augmentation was built, and the nighttime detection performance was improved.…”

Section: Related Workmentioning

confidence: 99%

A Data Augmentation Method Based on Generative Adversarial Networks for Grape Leaf Disease Identification

et al. 2020

View full text Add to dashboard Cite

The identification of grape leaf diseases based on deep learning is critical to controlling the spread of diseases and ensuring the healthy development of the grape industry. Focusing on the lack of training images of grape leaf diseases, this paper proposes a novel model named Leaf GAN, which is based on generative adversarial networks (GANs), to generate images of four different grape leaf diseases for training identification models. A generator model with degressive channels is first designed to generate grape leaf disease images; then, the dense connectivity strategy and instance normalization are fused into an efficient discriminator to identify real and fake disease images by utilizing their excellent feature extraction capability on grape leaf lesions. Finally, the deep regret gradient penalty method is applied to stabilize the training process of the model. Using a total of 4,062 grape leaf disease images, the Leaf GAN model ultimately generates 8,124 grape leaf disease images. The generated grape leaf disease images based on Leaf GAN model can obtain better performance than DCGAN and WGAN in terms of the Fréchet inception distance. The experimental results show that the proposed Leaf GAN model generates sufficient grape leaf disease images with prominent lesions, providing a feasible solution for the data augmentation of grape leaf disease images. For the eight prevailing classification models with the expanded dataset, the identification performance based on CNNs indicated higher accuracies, whereby all the accuracies were better than those of the initial dataset with other data augmentation methods. Among them, Xception achieves a recognition accuracy of 98.70% on the testing set. The results demonstrate that the proposed data augmentation method represents a new approach to overcoming the overfitting problem in disease identification and can effectively improve the identification accuracy.

show abstract

Section: Related Workmentioning

confidence: 99%

A Data Augmentation Method Based on Generative Adversarial Networks for Grape Leaf Disease Identification

et al. 2020

View full text Add to dashboard Cite

show abstract

“…FteGanOd + Faster RCNN) either outperforms or achieves similar detection rates on more complex datasets compared with existing methods on lower complexity datasets. The method of [22] achieved a precision of 97.1% and a recall of 55.0%. Our method (FteGanOd + Faster RCNN) achieve a higher precision of 97.8% at the same recall.…”

Section: Comparison With Existing Methodsmentioning

confidence: 97%

“…Most methods were evaluated using low-density traffic conditions and lowcomplexity backgrounds. The datasets we used (DATASET1 (easy) and DATASET2) contain an average of 5 target vehicles per image, and are more complex than those private datasets [1], [13], [14], [15], [24], [22]. Our method can adapt to different light conditions under natural driving conditions.…”

Section: Comparison With Existing Methodsmentioning

confidence: 99%

Feature Enhancement Based on CycleGAN for Nighttime Vehicle Detection

et al. 2021

View full text Add to dashboard Cite

“…With the development of GAN [53], more and more data augmentation technologies based on developed GAN have been proposed. In [54], a framework that converts daytime images into synthetic nighttime images based on a generative adversarial network was proposed, the generative adversarial network was trained on a public dataset and the experiment in a real nighttime dataset demonstrated that the performance of the model develops a lot due to augmented dataset. In [55], a generative adversarial network was used to generate real traffic sign images and the experiment demonstrated that the augmented dataset could develop the performance of the model, however, compared to GAN data augmentation, some traditional augmentation techniques could have a better performance which means that GAN is can be naively used for data augmentation but is not always the best choice.…”

Section: Methods For Data Augmentationmentioning

confidence: 99%

Driver Eye Location and State Estimation Based on a Robust Model and Data Augmentation

et al. 2021

View full text Add to dashboard Cite

Eye state evaluation is crucial for vision-based driver fatigue detection. With the outbreak of COVID-19, many proposed models for eye location and state evaluation based on facial landmarks are unreliable due to mask coverings. In this paper, we proposed a robust facial landmark location model for eye location and state evaluation. First, we develop an existing lightweight face alignment model for eye key point locations that is robust in large poses. Then, to develop the performance of our model in a complex driving environment such as an environment with mask coverings, changing illumination, etc., we design a method to augment the training data set based on the original landmark data set without any extra cost. Finally, some facial landmarks around the eyes are extracted, and the eye aspect ratio (EAR) is introduced to evaluate the eye state based on eye key points. The experiment shows that our model achieves significantly improved landmark location performance on a driving simulation data set due to data augmentation. We tested our model on the BioID data set to measure the eye state evaluation performance, and the results showed that our model obtained satisfactory performance with an accuracy of approximately 97.7%. Further testing on the driving simulation data set shows that our model is robust in different driving scenarios with an average accuracy of approximately 93.9%.

show abstract

Nighttime Data Augmentation Using GAN for Improving Blind-Spot Detection

Cited by 31 publications

References 29 publications

A Data Augmentation Method Based on Generative Adversarial Networks for Grape Leaf Disease Identification

A Data Augmentation Method Based on Generative Adversarial Networks for Grape Leaf Disease Identification

Feature Enhancement Based on CycleGAN for Nighttime Vehicle Detection

Driver Eye Location and State Estimation Based on a Robust Model and Data Augmentation

Contact Info

Product

Resources

About