https://store-images.s-microsoft.com/image/apps.46742.dbe533e6-fa9b-4a9e-bb72-fe252183b0e7.c2bda32e-4b1f-47a8-a509-1982d7ec854c.f7a2730b-863f-474c-8fda-b0976ef494e2

KMD Graphics Insight for Market Surveillance

KMD A/S

KMD Graphics Insight for Market Surveillance

KMD A/S

KMD Graphics Insight solution will find incompliant products on webshops based photos and texts

Online shopping is now so widespread that it would take tremendous human resources to screen all the items for sale online. Instead, the capacity required for screening, image recognition, and reporting hits to the authorities can be satisfied by artificial intelligence. Therefore KMD has developed Graphics Insight.

The idea originated when the Danish Parliament tightened the regulation on the field on 1 July 2020, granting the authorities a right to block websites if they continue to sell products that do not comply with Danish legislation and demands regarding consumer safety. At the same time, the sanctions are sharpened with higher fines and potential imprisonment.

Retraining, responsibility and ethics in AI

Although the AI in general is not considered a High-Risk AI in a High-Risk sector and the data that the AI is trained on does not contain personal information, it was evident from the beginning of the project that the solution should follow Responsible AI principles when using core machine learning technology.

It was important to build an AI solution with a Human-In-Command approach for all the scenarios where the AI is used. This helps to oversee the overall activity of the AI system and the ability to decide when and how to use the system in any situation. This may include the decision not to use the AI recommendation in a particular situation and to establish levels of human discretion during the use of the system and to ensure the ability to override a decision made by the system.

At the same time the authorities have planned audits for checking the quality of the solution, and the AI is constantly improved based on continuous feedback from the users. The ability to get feedback from the users in relation to learning from both false positives and false negatives is an essential part of the user interface.

https://store-images.s-microsoft.com/image/apps.64393.dbe533e6-fa9b-4a9e-bb72-fe252183b0e7.c2bda32e-4b1f-47a8-a509-1982d7ec854c.4a8a8ad8-a6d9-4d06-bd7a-52a0a83cc9b5
https://store-images.s-microsoft.com/image/apps.64393.dbe533e6-fa9b-4a9e-bb72-fe252183b0e7.c2bda32e-4b1f-47a8-a509-1982d7ec854c.4a8a8ad8-a6d9-4d06-bd7a-52a0a83cc9b5
https://store-images.s-microsoft.com/image/apps.3013.dbe533e6-fa9b-4a9e-bb72-fe252183b0e7.c2bda32e-4b1f-47a8-a509-1982d7ec854c.f28974ac-264e-4768-bf94-b2b17a3c23b6
https://store-images.s-microsoft.com/image/apps.46087.dbe533e6-fa9b-4a9e-bb72-fe252183b0e7.c2bda32e-4b1f-47a8-a509-1982d7ec854c.0e0cf921-2d48-45e7-920c-18a947c58b9d