Open emdo1 opened 5 years ago
Thanks for getting in touch about this. What sort of problems did you encounter?
I've never done anything with ultrasonic files – maybe have a look at https://github.com/macaodha/batdetect to see what they do about ultrasonic detection?
Thanks
I'm comparing the use of different acoustic methods, so I am using the bat_detect code to look at bats, but thought it would be interesting to see if CityNet could be used to infer overall biotic activity. I keep getting a ValueError: input signal length=0 is too small to resample from 384000->22050 when running without changes. I have tried changing the resampling to 192000 but that results is an error cascade.
Ok, interesting. I don't think I have enough ultrasonic experience here to be able to help you. My default first option would be to try to convert the ultrasonic files into files sampled at 48kHz, but obviously this isn't going to keep any ultrasonic signal.
As the models we have here were all trained on low-frequency files I wouldn't expect any kind of good results when applying them to high-frequency files.
Had problems running code on files with a sampling rate of 384kHz, struggled to find where to change settings for this. It wouldn't analyse the lower frequencies. However had no issue with 48kHz files.