Design
The designer uses Maestro to define the problem that the agent will resolve. No knowledge is necessary. Maestro will guide him from scratch to final package ready to deploy.
DATASET CONSTITUTION
A first step is to upload a small sample of representative records that will be used during the process. There are 2 usages of those records:
- For initial dataset constitution: even if iDetect-4.0 can start with an empty database it is often preferable to bootstrap it with an initial small dataset.
- For testing: those data are used to evaluate the performances of the agent before its deployment.
INPUT SOURCES DEFINITION
iDetect-4.0 is very versatile and can adapt to several sensor configurations. This task consists in defining the number of microphones/channels used (up to 4) and the kind of signal they are going to monitor (continuous or impulses).
QUALITY POLICY
The quality policy defines how the candidate models are qualified. It allows to give weight to specific results (False positives, False negatives) when computing the performance score of a model. This score will be used by intended service to take the decision to deploy the model or not. The way test datasets are constituted can be also finely tuned.
SPECIFIC DATA HANDLING
You can define which metadata coming from controller are stored along with the audio records. Those values can be, for example, a serial number, a temperature, or a pressure. Any value that can give you hints for determining the cause of a crisis.
SYSTEM SETTINGS
As part of the system settings, you can configure the UI look and feel, the different use credentials and the backup and cleaning policy.
