The following steps were completed Finetuning the YAMNET model from tfhub using available dataset Exported tflite model Using tflite in Android( Java) to get predictions ( Flutter also attempted) TODO Expand the current application to split audio file into 5sec interval, convert them to Java consumable format, predict each of them Then output the time of 'positively identified' sound activity Alternatively, while the java issue gets resolved, working on getting resmapled data from a flask API, the application will call this flask api.