Delair.ai platform enables to build and run in production Automatic Defect Detection apps to inventory and analyze categorized defects on power lines assets.
How does it work ?
- In a first phase an Artificial Intelligence algorithm is built and trained from your visual datasets and tested
- In a transition phase the algorithm is improved using an actual smaller dataset that is annotated and added to the initial one for an improved learning
- Then this algorithm can be used in production for any distance of power lines asset
The standard workflow employs image annotation, learning phase, and to run predictions.
What asset components can be analyzed ?
For example, and in base analysis, 4 categories of components are investigated :
Other categories can be defined in a customized approach.
What are the typically required inputs ?
For the initiation phase, we recommend datasets to be sized as following :
- Recommendation is of at least 2,000 pictures for each component , in exemple above 8,000 pictures for TSO (Transmission) power lines
- In a DSO case (Distribution), twice this quantity i.e. up to 16,000 pictures would be recommended given the higher complexity
Contact us to talk about the requirements and feasibility with your data.