jurek
Well-known member
I had an interesting thought.
Image recognition is used to everything from identifying faces of friends to cancer nodules. What if somebody develops a software which recognize and identifies birds on photos, and then it is paired with automatic CCTV cameras?
Imagine CCTV cameras set in reserves in remote beaches, pools or reedbeds, where people don't often visit. Or maybe automatic feeder camera recording birds when the owner is away from home. Or cameras in remote rainforests, islands or desert pools. Whenever something triggers a motion sensor, several photos will be taken, software would identify the bird and send a photo with a message via a mobile-like link.
Or imagine automatic camera stations left in the field for a year, with software automatically identifying birds and presenting a set of species for confirmation.
Software will not be foolproof. However, it is enough when a software recognizes at least there is a bird and sorts species with say 90% accuracy. This would sort pictures of bird species for a human to quickly accept or reject. This could do work which ornithologists could not process otherwise.
Imagine how much information about rare birds would be gathered. And how many rarities would be found and later twitched in places not well watched!
And if ornithologists might be afraid thier skills would become obsolete? Don't worry. Similar projects in other topics show that data grow so much, that although mundane tasks are automated, specialists have more work than ever.
Image recognition is used to everything from identifying faces of friends to cancer nodules. What if somebody develops a software which recognize and identifies birds on photos, and then it is paired with automatic CCTV cameras?
Imagine CCTV cameras set in reserves in remote beaches, pools or reedbeds, where people don't often visit. Or maybe automatic feeder camera recording birds when the owner is away from home. Or cameras in remote rainforests, islands or desert pools. Whenever something triggers a motion sensor, several photos will be taken, software would identify the bird and send a photo with a message via a mobile-like link.
Or imagine automatic camera stations left in the field for a year, with software automatically identifying birds and presenting a set of species for confirmation.
Software will not be foolproof. However, it is enough when a software recognizes at least there is a bird and sorts species with say 90% accuracy. This would sort pictures of bird species for a human to quickly accept or reject. This could do work which ornithologists could not process otherwise.
Imagine how much information about rare birds would be gathered. And how many rarities would be found and later twitched in places not well watched!
And if ornithologists might be afraid thier skills would become obsolete? Don't worry. Similar projects in other topics show that data grow so much, that although mundane tasks are automated, specialists have more work than ever.