|
|
|
|
|
|
by dcrimp
1030 days ago
|
|
|
Very interesting concept team. I know a fair bit about inspections for energy infra from my time at various utilities in NZ/AU and the UK - and I think you're on the right track. Would be interested to learn about your approach with computer vision for detecting defects. I worked closely with a project some time ago where we flew overhead lines with a drone. We had some issues with background separation - given the conductor hung at varying heights and the drone flew at a constant height, the camera was always refocusing - it was hard to get consistent results. It ended up resulting in false positives more than anything. I'm also interested in how you discern defects. Some sort of anomaly detection? ML? If so, how did you source training data? Also would be interested in how you determine speciation for underclearance of lines. So many cool elements of this to nerd out about. Congrats on launch, will be following closely! |
|
|
For now, we've focused a lot of our time on building the vector to gather the data. We did POC of code for real time power line following (so power line recognition and follow). We know that there are a lot of startups on this field of data analysis so we began scouting to see the level that is reached today. And decide if we should build our own model or use another one.