Demo of Cypress CY8CKIT-062-WIFI-BT connect with AWS IoT ConnectedCloudChallenge AI Code

Licensed under the Apache License, Version 2.0

Demo for element14 Connected Cloud Challenge project

LED and ALS sensor for sense the mail. Record Sound from TFT shield PDM MIC to on-board FRAM, stream via AWSIoT, pass the sound clip to SageMaker, run fine-tuning deep learning YAMNet to detect angry dog barking. Use AWS IoT shadow connected with IOS APP

More detail: https://www.element14.com/community/community/design-challenges/connected-cloud-challenge-with-cypress-and-aws-iot/blog/2020/05/24/deep-learning-for-angry-dog-sound-classification-11

Quick Install

Mbed and IOS reference of README.md under relative directory.

Location

1. IOS source ./ios-mailbox
2. Mbed source ./mbed-os-mailbox

Lambra python source

ambda.py

Build new lambda function under AWS console and paste it, set relative kinesis stream trigger.

test_sound_ai.py

Build a new lambda function under AWS console. follow https://docs.aws.amazon.com/lambda/latest/dg/python-package.html install python package requests and websocket

4. SageMaker Tensorflow

1. Under Amazon SageMaker, build a new Notebook instances
2. Git repositories https://github.com/tensorflow/models
3. IAM add IoT, S3 access
4. Open JupyterLab
5. File > New > Terminal
6. source activate tensorflow_p36
7. conda install -c conda-forge resampy
8. conda install -c conda-forge pysoundfile
9. conda install -c conda-forge libsndfile
10. copy iot_inference_yamnet.py iot_inference.py iot_train.py and angrydog.h5 to /home/ec2-user/SageMaker/models/research/audioset/yamnet
11. for fine tuning sound sample, copy dog and other folder to S3 Buckets soundsample, and copy inside test folder to S3 Buckets soundsampletest. Inside jupyterlab terminal > cd /home/ec2-user/SageMaker/models/research/audioset/yamnet > python iot_train.py

sound sample for training collected from freesound.org (Creative Common CC0) and testing from soundbible.com (Creative Common Attribution 3.0)

AudoSet Licence CC4.0 https://research.google.com/audioset/download.html

Our fine-tuning model reference of work by laanloabs: https://github.com/laanlabs/train_detector