Ordering clothes online just from looking at the items can easily end in disappointment. What you saw as red is actually quite pink, the zebra print is more of a leopard, and the large fits like a medium. Asaf Moses, 29, from Ra'anana, who has lived in Berlin for the past seven years, frequently encounters this problem. His fondness for secondhand clothes led him to do a lot of shopping on eBay, "but 80 percent of the time, I had to return the item." Then he thought of a solution. With no background in fashion, Moses and fellow Humboldt University economics student Sebastian Schulze launched a start-up called UPcload (a combination of cloth and upload ).

The basic idea is simple: Using a webcam, you photograph yourself from four different angles (the program tells you how ). You should wear snug-fitting clothing and hold an ordinary CD (so the distance from the camera can be estimated ).

The photo session takes about a minute. Then you receive body measurements according to a dozen parameters, including chest-waist-hips, as well as arm length, neck circumference and more.

Taking the data collected and cross-matching with the clothes sizes of different manufacturers, the program shows the consumer how each item of clothing, in any size, would look on him or her. Programs that assist in selecting clothes already exist, but they are based on measurements entered by the user, after taking them himself, while the UPcload model relies on measurements taken by the camera.

"When a person measures himself, there is a large deviation," Moses explains. "In the beginning, in order to compare it with our algorithm's performance, we measured by hand. We found that everybody has a different way of measuring, and that people aren't even aware of it. We brought in two professional tailors, and there were differences between them too. We invited people to try on clothes. We asked them to tell us from the outset which size they thought would fit them best. Only 60 percent got it right."

The company hired two employees whose job it was to be measured over and over again with the camera and monitor the consistency and accuracy of the results. Moses and Schulze also use a 3-D scanner at the College of Economics and Technology in Berlin to compare the results. So far, the deviation stands at 2-3 percent. "There are sophisticated 3-D scanners on the market already," says Moses, "but it's not something you can use at home."

The measuring algorithm is being developed in Israel by two experts in picture processing. All the rest - design, quality control and business development - is being done in Berlin. Humboldt University, which has also contributed funding, has allotted a floor of a building to the project. The German Finance Ministry has also invested in it, and in addition to prize money from winning entrepreneurial competitions and money from private investors, the company has raised 900,000 euros so far.

An open version of the program is due to go online in February (www.upcload.com ).