A Tool Chain for Active Shape Model

We present a tool chain to make the use of ASM algorithm in applications more easily. Active shape models (ASMs) are statistical models for deformable objects. It has been widely used for facial feature localization. But using the ASM is complex and requires a lot of work. This paper presents the design and implementation of a tool chain for the ASM based application. It includes labeling tool, training tool, fitting tool and test tool respectively. The labeling tool is used to label the features points in a face image and save the marked points to an annotation file very easily. The original code of training and fitting tools is from JIAPEI´s open source and we extend the fitting algorithm in our fitting tool. The test tool is used to divide the images into several groups by file names and apply cross-validation to test the accuracy of fitted feature points. Based on this tool chain, we get the fitting result with limited required effect. Finally, we introduce a gender classification prototype application based on accurate feature points detected by the tool chain.