Real Time Analysis of Twin Image Style Transfer
Hemanth Reddy Alavala, Sree Laasya Pabbisetty, S. Lalitha, Animela Niharika, M Shapna · 2025
Research in image styling has focused on developing techniques that allow the transfer of artistic styles to a content image, enhancing the flexibility and richness of visual representation. This research proposes a novel method that combines two distinct styles onto a single content image, in contrast to traditional strategies that usually employ a single style image. To combine two different styles, the VGG-19 model is used to extract features and create Gram Matrices. The final outcome may be fine-tuned using this method’s wide control over the blending ratio. According to the experimental data, the loss of style and content decreases significantly as the number of epochs increases. Additionally, the suggested method is quantitatively validated with a Peak Signal-to-Noise Ratio (PSNR) of 28.6 dB, which suggests minimal distortion in the generated images, a Structural Similarity Index (SSIM) of 0.92, which indicates strong structural preservation, and a Frechet Inception Distance (FID) of 45.3, which indicates high visual similarity.