Web Layout Generation And Optimization Based On Aesthetic Evaluation

Graphic design is popular in our daily life.However,many users are not able to produce high quality designs by lacking design knowledge and aesthetic ability.So it is significant to make designing graphic web much easily and more quality for novice designer.Layout design is one main topic in graphic design.Layout automatic generate and optimize methods now simply simulate traditional design principles to train a model by machine learning.The process is very complex and the results are uncontrollable.After the training is over,the system will be invisible to users.They can't see how the system works and affect the result,which usually lead to an unexpected layout.In this paper we introduce genetic algorithm to get an expected layout by multiple user selection.It separates the model training and layout producing system so the model can work as a fitness function individually to select layouts.And users can choose preferred layouts to control the route of layout evolution.Finally,layout generate and optimize are visible and controllable.We have built a simple frame for layout automatic generate,allowed to generate webpage layouts randomly.We also built an image analysis based model for evaluating webpage layouts.It is used as a fitness function to select layouts in the evolution process,which will ensure the quality of the layouts when users guide the styles of the layouts.