Data Augmentation on Synthetic Images for Transfer Learning using Deep CNNs

Talukdar, Jonti; Biswas, Ayon; Gupta, Sanchit

doi:10.1109/spin.2018.8474209

Cited by 17 publications

(15 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…NNs are a collection of neurons, simulate a network comparable to a human brain [ 11 ]. The NNs can adjust to changing input, so the model creates the leading conceivable result without overhauling the yield criteria [ 12 ].…”

Section: Neural Networkmentioning

confidence: 99%

“…A typical CNN [ 15 ] layers reduce the number of connections to improve computation with a lesser number of parameters as visualised in Figure 2.3 . Pooling is mainly done in two formats, maximum value pooling, and average value pooling [ 12 ]. The pooling layers discount the information to only the average of a kernel square or maximum of a kernel square.…”

Section: Figure 22mentioning

confidence: 99%

See 1 more Smart Citation

RCNX: Residual Capsule NeXt

Anilkumar

El-Sharkawy

2021

2021 IEEE International IOT, Electronics and Mechatronics Conference (IEMTRONICS)

View full text Add to dashboard Cite

Section: Neural Networkmentioning

confidence: 99%

Section: Figure 22mentioning

confidence: 99%

RCNX: Residual Capsule NeXt

Anilkumar

El-Sharkawy

2021

2021 IEEE International IOT, Electronics and Mechatronics Conference (IEMTRONICS)

View full text Add to dashboard Cite

“…Aiming to prevent overfitting, we used the technique of on-the-fly data augmentation, also known as online augmentation [30]. In our proposal, the applied transformations were the following: rotated ± 20°, horizontal and vertical translation of ± 3 pixels, and a reduction of the image size with a factor between 0.5 and 1.0 [31]. These operations resulted in an augmented dataset with 10,132 images.…”

Section: A Datasetmentioning

confidence: 99%

Applying Convolutional Neural Networks to Detect Natural Gas Leaks in Wellhead Images

2020

View full text Add to dashboard Cite

Detecting natural gas leaks is one of the most important measures in the oil industry for preventing accidents. The literature provides different techniques for detecting natural gas leaks. However, except for previous studies by the authors on this topic, there remains a gap in the literature on leak detection of natural gas using digital images, without the need for sensors or special cameras calibrated for the spectrum of methane molecules. These previous studies used image-processing techniques associated with a novelty filter classifier to detect the presence or absence of visible cloud of hydrocarbon vapors, that is, a natural gas plume in Closed Circuit Television (CCTV) frames installed in onshore wellheads. In this paper, we present a new method for detecting natural gas leaks in oil facilities that enhances the results obtained previously, along with the Gradient-weighted Class Activation Mapping Algorithm (Grad-CAM) to identify natural gas leaks. In this new method, convolutional neural networks (CNN) are applied to classify images (CCTV frames) as belonging to classes with or without natural gas leaks in onshore wellheads. Experimental results showed that the best performance model presented an accuracy of 99.78% and false negative rate of 0.00%.

show abstract

“…Some scholars tried to enhance and expand data by means of image processing algorithms [14][15][16]. The core idea is to regularize the target images in the original training set to generate new target images, without changing their labels.…”

Section: Introductionmentioning

confidence: 99%

Detection of Dense Small Rigid Targets Based on Convolutional Neural Network and Synthetic Images

Zhang¹,

Chen²

2021

View full text Add to dashboard Cite

Facing the image detection of dense small rigid targets, the main bottleneck of convolutional neural network (CNN)-based algorithms is the lack of massive correctly labeled training images. To make up for the lack, this paper proposes an automatic end-to-end synthesis algorithm to generate a huge amount of labeled training samples. The synthetic image set was adopted to train the network progressively and iteratively, realizing the detection of dense small rigid targets based on the CNN and synthetic images. Specifically, the standard images of the target classes and the typical background mages were imported, and the color, brightness, position, orientation, and perspective of real images were simulated by image processing algorithm, creating a sufficiently large initial training set with correctly labeled images. Then, the network was preliminarily trained on this set. After that, a few real images were compiled into the test set. Taking the missed and incorrectly detected target images as inputs, the initial training set was progressively expanded, and then used to iteratively train the network. The results show that our method can automatically generate a training set that fully substitutes manually labeled dataset for network training, eliminating the dependence on massive manually labeled images. The research opens a new way to implement the tasks similar to the detection of dense small rigid targets, and provides a good reference for solving similar problems through deep learning (DL).

show abstract

Data Augmentation on Synthetic Images for Transfer Learning using Deep CNNs

Cited by 17 publications

References 9 publications

RCNX: Residual Capsule NeXt

RCNX: Residual Capsule NeXt

Applying Convolutional Neural Networks to Detect Natural Gas Leaks in Wellhead Images

Detection of Dense Small Rigid Targets Based on Convolutional Neural Network and Synthetic Images

Contact Info

Product

Resources

About