ABSTRACT

The use of computers and the internet in today’s tailoring sphere ranges from communication between clients and their tailors to the marketing of products using modern marketing technologies and platforms. However, the cloth production process is still very manual and mechanical for the everyday small-scale tailor. From obtaining clients’ body dimensions to cutting of the fabric to the actual sewing process, it all requires the tailors’ manual input and for cases like body dimensioning, distance can be a barrier. This project aimed at computerizing the body dimensioning process and eliminating the distance barrier by developing a web-based non-intrusive body dimensioning system.

The clients were required to create a profile on the website developed as part of this project. They were instructed on how to take three pictures: a frontal picture with hands stretched out, a side picture with one hand stretched forward and another frontal picture where the clients hold a piece of paper with a circle printed on it. The circle is a hollow circle with outer diameter of 18cm which will be provided on the website. The pictures, upon upload by the clients, were fed into the python-based body dimensioning program which detects the circle on the calibration image – the image where the clients hold the A4 paper, obtains its outer diameter and calculates the ratio of the number of pixels spanning the radius to the known radius of the circle. The program implements BodyPix, a tensorflow-based person-segmentation AI model which segments the pictures and extracts the client’s body figures from the pictures. A mask of the client’s body figure is obtained from the side view picture and the client’s height is obtained by picking out the highest and corresponding lowest point in the body figure mask. Dimensions of other required body parts were obtained and converted to real-world distances using the ratio obtained from the circle.

The web-based non-intrusive body dimensioning system developed in the course of this project was deployed and tested on 9 individuals and was able to obtain their body measurements with an average error margin of 4.42%. The system was able to estimate nine body measurements namely: trouser length, shirt length, shoulder length, arm length, chest circumference, neck circumference, ankle circumference, wrist circumference and bicep circumference. The development of this system along with its internet presence shows that it is possible to eliminate the barriers of intrusion and distance faced by clients and tailors and gives the clients a wider array of options when choosing a tailor and also gives the tailors a much wider market reach.