Interactive Design-Assistance System Based On Machine Learning Methods

This thesis presents a brand-new interactive mapping system: MMC(Machine learning Mapping Creator)with machine learning methods,built with Max/MSP and font-end framework,is a universal and creative tool of interactive art creations.MMC achieves classification and identification of pictures captured by camera,uses the tags classified by the features to trigger events.MMC system provides a new way of making creative art,for those engaged in creative art making with machine learning methods,could perform well into different scenarios such as real-time art performances,new interface music instruments,installations,human-computer interactive applications and so on.When creating new interactive art creations such as real-time performances and installations,artists often choose to bring some advance technologies into the process,like sensors and some external hardware devises for acquiring real-time data streams,or complex models for generating music materials and so on.Those technologies contribute new ideas and methods to art creation,but also increase the complexity and difficulty during the creative process.The main purpose and content of this thesis is that figuring out how to bring advance methods such as artificial intelligence and machine learning techniques into the process of interactive art creation,and package these techniques into modules which are straightforward and simple to use.Users can activate the application built with font-end framework,set different tags for different gesture or objects.Then use the data which captured by the camera frame by frame to retrain the pre-trained model and get a new classifier using transfer learning method.The trained classifier can recognize which label the current gesture or object corresponds to.The tag message will be processed into a switch trigger patch in Max/MSP.Using these switches,users can make some mapping control with their own Max patch and some external modules or devices.This thesis mainly studies the scenario of interactive art creations that using the machine learning methods,and then general framework of an interactive system.Then gives some analysis of two related systems.Next,we describes the framework,implementation methods and functions of MMC system.Also we presents some practical ideas through some qualitative experimental analysis.Finally,the probable development of each modules in MMC is discussed.