A novel enhanced region proposal network and modified loss function: threat object detection in secure screening using deep learning

Deep learning neural networks for emotion classification from text: enhanced leaky rectified linear unit activation and weighted loss

“…The proposed customized LSTM model integrates dropout regularization, bidirectional processing, and weighted loss functions to deal with the data imbalance problem. The function loss can significantly reduce using regularization [3,37]. It outperforms the other options by 23.2% for the F1-score and 11.6% for MSE, respectively.…”

“…The goal is to minimize the cross-entropy loss [46], which can be defined in equation (3). However, the imbalanced data problem has not been considered, which can impact the overall average class accuracy.…”

“…With the rapid development of e-commerce and social networks, the Internet contains more and more usergenerated reviews [1][2][3] consisting of user opinions and emotions, which have significant business implications [4,5]. Because the emotions reveal user attitudes towards a product or service, they can be used for predicting sales, stock market trends, and more [6].…”

Automatic buildings detection using Sobel, Roberts, Canny and Prewwitt detector

“…The Canny edge detec-tor is the most time consuming in terms of calculations compared to other detectors published in this work. The computational complexity lies in the fact that the Canny edge detector requires two mathematical parts, namely the convolution of the original image with a Gaussian filter (9) and then the convolution with the Sobel operator (1). Despite the computational complexity, we decided to implement the Canny detector in our software solution, due to its favorable results.…”

“…In the field of medicine, it is possible to use AOR for the DNA damage analysis, or in the Mycobacterium tuberculosis identification system. Another possible application of AOI is its use in the detection of insulators in the power system [1][2][3][4][5][6][7][8].…”