GCP for Data Scientists
summary
A project showing how Google Cloud Platform speeds up a Raspberry Pi baby-cry detector: audio is split into five 1.8s segments, features extracted with librosa, and a 3-of-5 majority vote decides detection. Running prediction locally on the Pi takes ~45 seconds, while uploading audio to Cloud Storage and invoking Cloud ML Engine via a Cloud Functions webhook returns five model predictions and the final vote in under 5 seconds; deployment details include enabling GCP APIs, a functions requirements file, logging, and an estimated small test cost of about $0.08/month.
AI-written summary · about
chapters (25)
- 0:00Opening
- 0:45Context
- 2:33Democratizing AI
- 4:29What I understood about the cloud
- 4:55The different levels of cloud computing
- 8:46Why the cloud?
- 10:00What is the cloud?
- 10:21Available services
- 12:47Training
- 13:32The 3 concepts
- 14:20Practical example
- 15:53Training vs prediction
- 16:12The potential of the cloud
- 17:25Predictions on ready-made data
- 18:11Audio predictions
- 19:28Basic script
- 21:17Cloud Functions
- 22:40Predictions
- 23:18Retrieving predictions
- 24:19Prediction results
- 24:56Code presentation
- 25:41Testing the functions
- 26:45Calling CloudML
- 29:36Next steps
- 31:36Conclusion
related talks
15mGCP for Data Scientists
53mCloud Machine Learning APIs for fun and profit with Harry Potter
44mAdventures in Puppy Training with A.I. and a Raspberry Pi
59mMachine Learning and Big Data Magic with Node.js and Google Cloud
07mOk Google, What Am I Doing? Acoustic Activity Recognition Bounded by Conversational Assistant ...
40mPuppies or Muffins
22mAnalyzing Pets and Live Debugging: Using Google's Machine Learning APIs and Stackdriver Debugger
44mBuilding smarter solutions with no expertise in machine learning