ML627170905
Contributor
Date
Location
- Age
- Not specified
- Sex
- Not specified
- Playback
- Playback not used
Observation details
Noes on AI / Machine Learning, in this case specifically birdNET. Prior to this trip there were only 30 recordings of this species in MacAulay, suggesting data deficiency for identification training in Merlin and birdNET. As is often the case when encountering “untrained” species, birdNET returns multiple possibilities within the same genus, with variation depending on the particular audio segment analyzed. For example, for one phrase of this recording birdNET proposes Arremon flavirostris, Saffron-billed Sparrow, at its highest level of certainty, “Almost certain,” at the same time suggesting Arremon schlegeli, Golden-winged Sparrow, at its next level of certainty, “Highly likely.” (Important to note that I use the application post-facto with the location information disabled, so that the application is forced to compare recordings to the entire world database of species—possibly on location with location access turned on, it might be trained to successfully interpret Arremon vocalizations in Salta / Jujuy as Moss-backed Sparrow—I don’t know.)
Additional species
Technical information
- Recorder
- Microphone
- Accessories
- Original file size
- 12.75 MB