tsarprince/dog-bark-detection

Dog Bark Detection

Conda Environment

conda env create -f environment.yml
conda activate dog-bark

Basic Exploration

If you're just looking to play around with the ML model without worrying about the AWS cloud components I'd strongly suggest looking at notebook.ipyn after running the conda commands. It's set up to train the model from scratch and run inference over .wav files.

Run Prediction

QUEUE_NAME=<queue_name> TABLE_NAME=<table_name> python3 main.py

Docker

docker build -t dog-bark .
docker run --env QUEUE_NAME=<queue_name> --env TABLE_NAME=<table_name> dog-bark

ECS

# Login
aws ecr get-login-password --region ap-southeast-2 | docker login --username AWS --password-stdin 123456789012.dkr.ecr.ap-southeast-2.amazonaws.com/dog-bark-detection

# Build
docker build -t dog-bark-detection .
docker tag dog-bark-detection:latest \
    123456789012.dkr.ecr.ap-southeast-2.amazonaws.com/dog-bark-detection:latest
# Push
docker push 123456789012.dkr.ecr.ap-southeast-2.amazonaws.com/dog-bark-detection:latest

Kubernetes

gcloud auth login
gcloud config set project <PROJECT>

docker build -t gcr.io/<PROJECT>/bark-detector:latest .
docker push gcr.io/<PROJECT>/bark-detector:latest

# When connected to Kubernetes
kubectl apply -f deploy.yml

Attribution

• Sound Classification using Deep Learning