Birda supports AI-based bird photo identification for species in North America, Europe and Southern Africa.
Are all species covered for each region?
Birda's AI model is training is primarily prioritised based on how common each species is in a given region and then by the number of photos we have of each species.
North America: 99% coverage of the top 700 most commonly seen North American species and 41% coverage for the remaining least commonly seen North American species.
Europe: 98% coverage of the top 700 most commonly seen European species and 30% coverage for the remaining least commonly seen European species.
Southern Africa: 88% coverage of the top 700 most commonly seen Southern African species and 34% coverage for the remaining least commonly seen European species.
Is Birda working on improving coverage in North America, Europe and Southern Africa?
Yes, we will be making regular updates to the AI model as we get more training data for species without coverage.
Is Birda working on improving the performance of the AI model?
Yes, we will be making regular updates to the AI model based on species performance reporting. So when we see that the model is now performing well for a specific species, we will review our training data to improve model performance.
Why is the location and date required to get an identification?
Birda requires the date and location of the sighting for two reasons. The first is to improve the accuracy of suggestions, so please make sure you use an accurate location and date. The second is so that you can immediately post the sighting to Birda after getting an identification for your photo.