From eBird:
Are you a photographer? Share your images to advance bird ID research!
August 18, 2012
Merlin
What if a computer could identify the Merlin in this photo? With
recent advances in computer vision, this idea is no longer a pipe dream.
The Cornell Lab is teaming up with computer vision researchers, the
Visipedia team, at UC-San Diego, UC-Berkeley and Caltech to tackle this
challenge. To get this effort rolling, we are building a database of
70,000 correctly identified images to train our system to identify 550
North American birds, and we are looking for people to share their
images of these species. Click here to read more!...
Image Guidelines: To train our computer vision algorithms, there are a few photo characteristics that are most useful. Strive to meet these criteria with the images you upload:
• Single bird takes up most of the frame
• Image includes the entire bird (wings, tail and legs are not cut off)
• Color of the bird appears as they do in the field. Avoid images that are over-exposed, over-saturated or with heavy flash use.
• Bird is not too obscured. Complicated backgrounds are ok, but a bird that is mostly visible is important.
No comments:
Post a Comment