Advancements in Artificial Intelligence have spurred a never-ending list of practical applications whereby AI is used to boost productivity. Bread AI is just one mere example.
Launched in 2013, BakeryScan is an AI that was produced in order to differentiate between hundreds of bread types and output their prices in an instant. The bakery that implemented this technology had several types of bread, yet they could not be wrapped individually, as customers would reject them due to the assumption that they were not fresh. This meant that price labels could not be put on the bread, and the staff at the cashier could not memorise every single type of bread and their prices. Therefore, BakeryScan was developed so that the bread could be identified just by scanning it. The AI played a large role in reducing waiting time while buying bread, helping both the staff and customers.
How AIs Learn
Nowadays, most AIs function via ‘deep-learning’, which means that their databases are built through layers. They start off identifying basic borders and get increasingly complex, eventually analysing entire shapes and faces once supplied with enough varied data. Bakeries that develop different types of bread rapidly may find it different to supply such a large amount of data every time it is required. Hence, BakeryScan relied on individual hand-tuning to input examples of each bread and make algorithms for the way lighting and bread that was too close together or torn apart affected how the bread was identified.
Further Developments
In 2017, a doctor realised that bread was slightly similar to cells under a microscope, and contacted the company that developed it, BRAIN, to see if they could provide an AI that could identify cancerous cells as well. The core program within the AI was named AI-Scan, and this was the framework for the program that would identify the cancer cells, Cyto-AiSCAN. AI-Scan was later adapted to be used in a variety of unrelated situations, such as determining the number of people featured within centuries-old woodblock prints.
Sources
Second Image Source: https://rzksenpai.blogspot.com/2021/03/99-accurate-japan-create-bakeryscan.html
Written by Mana Ganesh Kawamura
Edited by Nichapatr (Petch) Lomtakul