Hacker News Reader

Comment by IndySun | Hacker News Reader

These apps do not work. I've tried all of them.

Every app listed here, and the OP BirdNet, are so ineffectual. They have no idea what they are listening to (unless location is on) and they cannot discern a genuine tweet from a washing machine squeak. The posters saying the apps are verging on incredible are completely deluded. I have used these apps, all of them, across Europe and North America, and they are inept and fail at a rate of 100%.

Who are you people?

And the top remark from the person that 'had this idea 10 years ago'... is this a joke?

Where is YC going?

View on HN

I whistled to the app and it identified the sound as human. I haven't gotten a chance to try it with an actual bird yet.

Location (as well as date, time, light, air pressure...) is certainly a factor that should be used when classifying something though.

I think the problem is probably related to data collection / data integrity. At least in my case, I'm using an iPhone that has variously placed microphones and speakers. It isn't clear to me how to hold the tool to collect the sound. If it gets wet or grungy, it's not something that is user-serviceable. I don't have an easy way to tell if it even needs to be serviced. If you're relying on the microphone that follows you literally everywhere to remain clean & untarnished - maybe that's the problem?

My phone rides in my pocket, it picks up dust, it gets moist, it gets dropped, etc. Perhaps I'm projecting. Maybe your phone is untarnished and is working fine? Do you have data to back up the 100% fail rate across Europe and North America? It seems you might be projecting as well?

I work with audio, sometimes bird sounds (for interactive encyclopaedias), playing the driest, clearest, sound into the app is often right, but not always. You make a good point about the microphones. Sweeping through a sinewave can broadly tell you something about what the phone mic 'hears'. Most city birdsong ambiences are very complex soundscapes. There's some important characteristics in the higher frequencies of birdsong that will not be 'heard'.

GPS helps to narrow down the the potential bird choices. So over time apps will learn and become more accurate. The open source Spleeter technology, unmixing stereo music, and its ilk, may be helpful run in real time allowing the app to ignore sound that is not birdsong.

Reply on news.ycombinator.com