
On December 19, 2019, Takashi Someda, HACARUS’ Chief Technical Officer, Naoki Kitora, its Chief Data Scientist and Ryuji Masui, Data Scientist gave a lecture at Shiga University, to the Graduate School of Data Science, introducing real world examples of utilizing machine learning for improving visual inspection at manufacturing sites.
The Graduate School of Data Science at Shiga University, which opened in April of this year, has a mixed student body, consisting of professionals, in addition to those who have advanced directly from undergraduate degrees – and their backgrounds are wide-ranging, including manufacturing and IT. Taking this into account the lecture focused on presenting a business driven, problem solving approach to machine learning, and how HACARUS applies techniques to solve real world problems.
During the lecture, based on a wealth of knowledge from customer engagements, HACARUS outlined and discussed key issues and obstacles that needs to be addressed when building machine learning based solutions for manufacturing use cases, these were:
- Little data on defective products
- Defective products cannot be overlooked
- Ambiguous definition of defective product
- Severe time requirements
- Data security policies
- Accountability
- False expectations
HACARUS introduced real world examples of lessons learned from projects, including what was being tested, what the performance requirements were, and what could be done with existing imaging devices and lighting. At its core, data science is, when applied to business, not merely about algorithms but equally about successfully capturing customer issues.
During the hands-on training part of the lecture, rather than simply using CNN to perform image analysis, a framework based on first understanding the data sample was introduced, and that good results can be had, even when using less complex methods. This part of the training was meant to give students new tools to solve issues, rather than using a one-size-fits-all approach.
This business driven perceptive was appreciated by the students, with one reaching out following to share the following feedback:
“The story of HACARUS who practices data science was valuable and very informative. Especially in solving business issues with data, I realized that communication with customers and other departments that actually use them was important. I will remember to keep track of customer’s challenges.”
In conclusion, at HACARUS we believe that it is indispensable for service providers and users from various perspectives to understand the characteristics of AI development projects and gain insights into them.
By combining understanding of the technical aspects and business realities the AI solutions of the future will be made. We are happy to give lectures at educational institutions where the next generation of data scientists are trained.
If you belong to an educational institution and have an interest in a HACARUS guest lecture or workshop, please contact us here.