TensorFlow Hub for real world impact
summary
TensorFlow Hub supplies ready models and shows concrete wins: an audio model flags more than 1700 gunshot instances in 350 GB of data in under 15 minutes; swapping feature vectors boosts a bean-disease classifier from ~48% to ~60% to 96% accuracy when using a cassava-trained feature vector; model pages list architecture, dataset and license and some have browser demos; many models export to TF Lite on-device deployment so classifiers run offline on phones.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:56Model Page
- 7:15Monitoring Forests
- 7:30Acoustic Analysis of Data To Support Wildlife Conservation
- 9:03Transfer Learning
- 10:37Build the Model
- 14:17Tensorflow Forum
related talks
21mTransform your web apps with Machine Learning
30mTensorFlow for Mobile Heroes
06mPyTorch 1.2 and PyTorch Hub: A Quick Introduction
38mThe Hitchhiker's Guide to Tensorflow: from Tensors to running models in production
01h 09mReinventing Machine Learning with Transformers and Hugging Face
33mMachine Learning 2.0 with Hugging Face
31mTensorflow for Janitors
41mTensorFlow ever higher