Automatic Self Calibrated Measurement System Using 3D Modelling

The growing era of digitization and automation of the apparel industry offers massive opportunities to improvise upon the traditional methods. The design process is a fundamental and preliminary action amongst a series of merchandise manufacturing processes. The designs obtained via customer interaction must be accurate and precise enough to acquire a satisfactory end product. To achieve the required precision of the perfect fit attires at an affordable price, one must be able to measure themselves from time to time and hence need a quick and comprehensive tool to do this with ease. In this paper, an approach is proposed that automates the measurement process bypassing the human involvement of the tailor to obtain measurements swiftly. An object is detected from the input followed by key-point detection and instance segmentation. After instance segmentation, a 3D model is generated using deep learning and machine learning techniques like Convolutional Neural Networks, 3D Reconstruction. The system takes input as an image or a video, analyses the 2D and 3D forms of measurements from the input, and finally outputs the required set of precise measurements from the respective input. The results obtained by this method are more accurate compared to traditional methods of measurement.