Post:
Update:
DataSource:
TextView:

databaseName: BirdSongs.db
cnnModelName: model_raw.pkl

Birding Via Mic CNN

This version is BirdingViaMicCNN
Identify bird songs using the Convolutional Neural Network process

Missing: BirdingViaMicCNN Image

You can drop a bird song here and click 'Play'.
Or you can click 'Play' for a random selection.
Then you can click 'Id'.
It will be transferred to the model to be (hopefully) identified or ckick play to play a random song.
Or you can click 'Id' in the Android app BirdingViaMic and it will go through this web page to the model and the results will be returned to your app.

Random Sample Test

   

Overview

https://www.modelsw.com/Android/BirdingViaMic/BirdingViaMic.php
Gerard Gilliland MODEL Software, Inc.
This version is BirdingViaMicCNN
Identify bird songs using the Convolutional Neural Network process
gerardg@modelsw.com

based on speaker_id.py
performs speaker_id experiments with SincNet.
Mirco Ravanelli
Mila - University of Montreal
https://github.com/mravanelli/SincNet

populated using bird songs from https://www.xeno-canto.org

build database in SQLite from IOC World Bird List by Frank Gill.
https://www.worldbirdnames.org/new/

The model is on NVIDIA
using Convolutional Neural Network model built on Ubuntu 20, Python 3.6, Nvidia Quadro GV100 GPU
run using terminal$ python3 species_id_prediction.py using the GPU
For Android converted to use only CPU.

I started out running python model on Android
I could run locally loaded from Android Studio using
https://chaquo.com/chaquopy/license/
https://chaquo.com/chaquopy/doc/current/examples.html
https://www.youtube.com/watch?v=dFtxLCSu3wQ
the complete Android Birding Via Mic CNN is on Github:
https://github.com/gerardgilliland/BirdingViaMicCNN
and the source code is under C:\AndroidProjects\BirdingViaMicCNN
this app BirdingViaMicCNN (Convolutional Neural Network) is currently standalone
I planned to integrate it into BirdingViaMic -- but I am not there yet.
this runs but can only handle short songs.
But all of the above doesn\''t matter -- Google Play store would not accept the model size.

So this is plan B
C:\OSG\Accounts\Android\WebPage\BirdingViaMicCNN
send an unknown song from exising Android Birding Via Mic to this web site
https://www.modelsw.com/Android/BirdingViaMicCNN/BirdingViaMicCNN.php
run the model here to identify (or try to identify) and send the bird species back to android.
Currently, you can click only click Play and a random selection of the 6 files will be chosen.
Then click Id to identify it using the model.
I will respond with a Bird Name to a file sent from Birding Via Mic (after this works and I send one).




mailbox
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
You may send an email: