Belezati: desenvolvimento de um módulo de escaneamento e descrição de ingredientes cosméticos baseado na web utilizando tecnologia OCR via API

Abstract

A survey carried out in this work with 254 participants revealed that the high price of a cosmetic product is a determining factor in the participant not purchasing it. But, the participant could change their mind and purchase it upon learning the product benefits can provide. We can conclude that the value of a cosmetic product is not related to its prices to the cost-benefit it provide. In this sense, we designed a way to help consumers make better choices when purchasing cosmetics, no longer choosing based on the product price but preferably based on the benefits it can provide. In this way, we develop a module within the existing web system on the market, called Belezati, to inform consumers about the function of the ingredients contained in the composition of a cosmetic. One way to offer this functionality in a practical way when purchasing a product would be the possibility of scanning the ingredients on the cosmetic packaging and instantly receiving information about the product. For this, we implement two optical character recognition tools available on the market through APIs. The tools chosen were Tesseract and Amazon Textract. After the analysis, we conclude that the Amazon Textract tool has a higher performance than Tesseract, which requires prior image processing to extract the text. Another success factor for extractions is the image resolution, which, the higher it is, the better the sharpness of the characters, making extraction easier by the tools. In addition to perceived performance, during the experiments it became clear that tools that use optical character recognition help people with low vision understand what on cosmetic labels is written, which often contain characters too small to be read by eye.