Faster deep learning solutions...
summary
An end-to-end Intel Deep Learning SDK installs Caffe/TensorFlow and MxNet, runs in Docker containers on Windows/Mac/Linux or AWS, and exposes a web UI plus Jupyter notebooks for dataset prep, live training graphs, and testing. Model optimizer does quantization and compression and exports an intermediate format for an inference engine targeting CPUs and FPGAs. Claims show typical CPU performance gains of 10–20× versus vanilla frameworks; demonstration trains a map-tile classifier (swim vs not swim) with live feedback and interactive testing, and the pipeline successfully produces deployable, optimized models.
AI-written summary · about
chapters (10)
- 0:00Start
- 1:36deep learning today
- 2:12DEEP LEARNING Deep neural networks are solving real life cognitive tasks
- 2:53Deep learning is Everywhere
- 4:41DEEP LEARNING steps Step 1 Training
- 6:14Intel® Deep Learning SDK - Workflow
- 7:21Plug & Train - An easy to use
- 11:50Model Optimizer
- 12:46Inference Engine Features
- 30:21Download, use, and provide feedback
related talks
24mTools & Techniques for a cost optimized Analytics & AI pipeline
58mDeep Learning in Microsoft Azure: CNTK, CaffeOnSpark and Tensorflow
35mAccelerate deep learning in the cloud with custom ML environments
44mDeep Learning in Production
37mDeep Learning Neural Network Acceleration at the Edge
57mDeep Learning with C++
28mDeveloper Essentials for the New Era of AI
18mDeploying ML solutions with low latency in Python