Euremo required a method for gathering massive amounts of data for market research. The study's goal is to automate the data collection process by gleaning information from the labels of fast moving consumer goods.
The platform's main purpose is to help scientists work more effectively. The importance of maintaining quick transactions and providing a mechanism for them was driven home after observing a researcher at work on a typical day.
Numerous variants of the FMCG items are available. Due to the wide variety of product packaging, it is impossible to set any defaults in advance. This led us to gather representative samples of roughly 150-200 products in a wide range of colors, fonts, languages, surfaces, etc. The next step involved training the models to recognize a large enough sample of unique wrapper types to be statistically significant.
Another aspect of the issue was figuring out how to make it so a small central admin team could efficiently correct the data on a massive scale. The solution was to create a web-based admin through which the data could be mapped onto the image and the data pool could be quickly scanned.
Supporting any possible data source is crucial to the success of such a research program. Without limiting the users with the need to control the input format, the Emproto team created a system that can read the data through OCR and tabulate in a comprehensible, analyzable format.
Remote Visualization - This helps user to just walk into a store or even be at their home, but visualize the future as it is, from anywhere
WYSIWYG - What you see is what you get. This app has been made with utmost effort to do the same
Researchers with a large data collection task were the intended users of the platform. The main obstacle was making sure everything ran efficiently. - Maximum information gathered in the shortest possible timeData correction mechanism for massive datasets
Generalised algorithm to tabularise the data