Related Experiment Video
Updated: Oct 21, 2025
04:17
DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning
Published on: May 10, 2024
1.0K
Neural Network-Based Mapping Mining of Image Style Transfer in Big Data Systems
Hong-An Li1, Qiaoxue Zheng1, Xin Qi2
1College of Computer Science and Technology, Xi'an University of Science and Technology, Xi'an 710054, China.
Computational Intelligence and Neuroscience
|September 3, 2021
Summary
This study enhances image style transfer using VGG-19 networks with L1 and perceptual losses. The improved deep learning model effectively preserves image content while achieving better style transfer results.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Machine Learning
Background:
- Image style transfer is crucial for big data applications, utilizing neural networks for image data mining.
- Deep learning methods for style transfer often result in the loss of original image content information.
Purpose of the Study:
- To address content loss in deep learning-based image style transfer.
- To improve the model's perceptual ability and maintain image content during style transfer.
Main Methods:
- Incorporated L1 loss into the VGG-19 network to minimize discrepancies between image style and content.
- Introduced perceptual loss to compute feature map semantic information, enhancing model perception.
Main Results:
- The proposed method significantly improves style transfer capabilities while preserving image content.
- Evaluation metrics showed increases in structural similarity (0.323%), cosine similarity (0.094%), and mutual information (3.591%).
Conclusions:
- The enhanced model provides superior stylization that better aligns with user expectations.
- The integration of L1 and perceptual losses offers a robust solution for content-aware image style transfer.
