Researchers at Columbia Engineering, have developed Birdsnap which is an iPhone app that makes bird identification very easy.
The programme lead by Computer Science Professor Peter Belhumeur, relies on an electronic field guide featuring 500 of the most common North American bird species.
The app is free and helps users to identify bird species through uploaded photos. The website at present has about 50,000 images.
Replying to Green Clean Guide’s queries, Thomas Berg, a Columbia Engineering computer science PhD candidate who works closely with Belhumeur said, “for now, I’m afraid the app is iPhone-only, although as an Android user myself, I’m trying to make an Android version happen. Whether we expand coverage will depend on the reception of the initial version–we’d love to do it if interest is high.”
“As for the accuracy of the software, it’s difficult to quantify as the app has just launched, and we’re not always able to tell whether the results from the app are correct (we are a computer vision research lab, not ornithologists or expert birders). We have done some controlled tests, to be published at an academic conference later this month, in which we get the correct species in the top five guesses about 80% of the time, however these were mostly pretty high-quality images; the images uploaded by our users are much more variable”.
I think the two most important limitations are the area of coverage, and this sensitivity to image quality. I hope we’ll be able to improve on both counts!” he added.