When you work with Image Recognition, you are expected to do magic. You upload the image to the app, and in a few seconds, you're getting the list of SKU that is seen on the shelf.
To roll out technology like this, companies have two options:
1. Train neural networks to work automatically
2. Hire outsourced teams that will do recognition manually.
Why does it matter?
Apart from the price, which is high when you need to use labor, the value of image recognition in non-real time is lost. Sales representatives are not getting any call-to-action reports and leave the store.
How could you know that your vendor uses people in the loop for Image Recognition (using it for neural network training is OK, btw)?
Try googling company reviews at Glassdoor, and you can get results like that:
“The solution is far from automatic while we try to add automation, we keep adding manual teams. teams are stuck.”
“ The service is a scam, you tell the clients that the image recognition is AI-based with machine learning however, in fact, human operators in Tunisia and Madagascar do the image recognition manually.
- as a service owner, the person who keeps in contact with clients needs to find out "stories" to explain when recognition errors happen. Although many times, it's a basic human error due to, e.g., lack of training, the explanation is like: "we had to fine-tune our AI recognition algorithm, our cloud service provider made a mistake, etc.”
Another way to check it yourself - visit Similarweb.com and type in the website name to see the audience. A company in the US shouldn't have 90% of its audience from Sri Lanka. Below is a screenshot of such a report.
