Efficient balanced focal loss function for manipulated images detection
Fatima Zahra El Biach, Imad Iala, Hicham Laanaya, Khalid Minaoui · 2021
Convolutional neural networks based segmentation methods of falsified images are confronted with the imbalance class problem; Imbalanced classes generally refer to a classification problem where the observations ratio of a class to all observations is very low. This class imbalance clearly increases the difficulty of learning and introduces strongly biased predictions in favour of high precision but low recall.In this article, we propose the balance factor based a focal loss function to solve this data imbalance problem. The experimental results show that the proposed method presents the best performance in terms of Accuracy, MCC and F-Measure on the CASIA-v1 and CASIA-v2 database.