Development
How it's working?
Phase I: Gathering the data
Process of training the artificial intelligence model based on the collected data.
Problem Identification
What kind of categories should AI recognize? Examples:
- Good vs damaged items
- Fresh vs gone bad
- Different material classes
- People identification
Documenting
Gathering named photo sets can be managed through a special mobile application that can be easily installed on the smartphone.
Every photo needs to be properly categorized by choosing it’s category manually.
Archiving
After documenting, each photo is sent to the archive in the cloud where our algorithm will use them to learn to classify chosen objects.
>How it's working?
Phase II: Training
Process of training the artificial intelligence model based on the collected data.
Choosing the right architecture of “splot” neural network
Choosing the right CNN architecture is crucial to be able to get the best results to recognize images.
Adjusting model’s hyperparameters
Choosing proper hyperparameters to a given CNN architecture and adjusting them to get the best results.
Testing on data that wasn’t used for training
This allows to test in the conditions that will be later present later while being able to control the results and adjust steps 1 and 2 if needed.
How it's working?
PHASE III: Deployment & integration with existing solutions
Deployment process of the trained model and integration with existing systems and devices
“wystawienie modelu” in the cloud or on the edge devices
The model with it’s API can be Choosing the right CNN architecture is crucial to be able to get the best results to recognize images.
Administration portal
In order to control access, billing, configuration of basic parameters, admins have the ability to manage the mechanism through a web portal.
Integrations
System can communicate with a any chosen software, thanks so the ability to create dedicated integrations.
