What's New with TensorFlow?
summary
Shows concrete releases and numbers: TensorFlow Lite ships a .tflite runtime that runs MobileNet image inference on a Pixel 2 in ~50 ms (emulator ~1,100 ms); TensorFlow.js runs training and inference in the browser and Node; Cloud TPUs push ResNet training to ~2,700 images/sec and reached 76% ImageNet accuracy for about $85 in those benchmarks. AutoML / “learn to learn” networks beat hand‑designed MobileNet/Inception at the same compute cost, and diabetic retinopathy models match or slightly outperform top ophthalmologists.
AI-written summary · about
chapters (11)
- 0:00Start
- 13:21Different scenarios for machine learning
- 16:49Using TensorFlow Lite
- 18:14TensorFlow Lite performance
- 25:08Training loop- eager execution
- 26:01Define models
- 28:09Input processing
- 33:01Launching new resources for learning and sharing
- 33:42Empowering everyone to learn and explore ML
- 34:16TensorFlow Hub
- 38:06AutoML outperforms handcrafted models
related talks
32mAI for Mobile and IoT Devices: TensorFlow Lite
14mOptimize your TensorFlow Lite models
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond
48mTensorFlow.js
34mTensorFlow Lite
41mTensorFlow.js 101: ML in the Browser and Beyond
21mA Journey into the World of Machine Learning with TensorFlow.js
38mReal-world Machine Learning with TensorFlow and Cloud AI