Design of Semantic Segmentation Algorithm to Classify Forged Pixels
Abhishek Thakur, Shamneesh Sharma, Tushar Sharma · 2023
Anyone can use editing software to rotate, scale, or compress photos without permission this creates the challenges of authenticity and integrity. In this paper, a deep learning-based algorithm is proposed. In this paper semantic segmentation is used to classify forged pixels. In this algorithm, ninety-one layers are used. This model contains five blocks of encoder and decoder which contain convolution, ReLU, and pooling layers. The copy moves and splicing image CASIA dataset is used. The model is trained with sixty percent of training images and forty percent of testing images are used. This model is trained on NVIDIA 1070 GPU for 1810 minutes with 33292 iterations. This model produces an average accuracy of 98%. The model is compared with other methods, and it shows better accuracy than others.