A Creative Idea Stimulation Method Of Image Based On Deep Learning Technology

Creative idea generation is the core part of the design process.An effective creative idea stimulation method plays a significant role in innovative design.However,stimulating creativity is often an abstract and elusive step in the design process.In the big data era,many image materials with potential creative stimulation abilities are still asleep.They do not provide power in promoting the idea generation process for designers.Currently,deep learning technology has achieved superior performance in the field of large-scale image retrieval,video analysis and image generation.By adopting the deep learning technique,it will bring great opportunities for innovative design by connecting large-scale image materials with creative idea stimulation.In the context of a sharp increase in image data,our study aims to propose a theory and method to stimulate creative ideas for designers.By making full use of the design process data,combining deep learning techniques,and solving the problems of inappropriate use and weak creative abilities of current image materials,we can guide the innovative design process.First,we explored the opportunities of the image materials' stimulation in the whole design process.By decomposing the design process continuously and using the methods that combined qualitative evaluation and quantitative evaluation,we could compare the performance of the designers at different design stages.It is concluded that the initial stage of thinking convergence in the conceptual sketch design process is the best stage for applying image stimuli,that is,the stage that has better absorption and transformation effects on images without negative effects.Meanwhile,we established a creativity evaluation system for sketches.Second,we carried out the eye-tracking experiments to study the image selection behavior of designers when improving their design works.we analyzed the connection and difference between the subjective selection results and the performance of the eye movement signal.Using the characteristics that they are closely related and complemented each other,we established an image material creative stimulating ability evaluation model to mark the creative stimuli ability of mass image materials.Third,we used convolutional neural network(CNN)to extract image features,canonical correlation analysis(CCA)to establish the connection between the input image to be retrieved and the output image with high creative stimuli ability to achieve image retrieval.By evaluating and feedbacking the search results,the deep learning model was optimized.Thus,we built an intelligent image retrieval mechanism and completed the transformation of massive image data to highly creative images which are useful to designers.The study proposes a method that stimulates creative idea with image materials based on deep learning technology.It combines the designer's design thinking and artificial intelligence technology effectively and accelerates the innovative design process significantly.It provides a new way to enhance design thinking,and new ideas for the development of intelligent design.