Introduction to the Microsoft Cognitive Toolkit (Deep Learning Framework)
summary
CNTK is a production deep learning framework that trains on CPU, GPUs and across machines using MPI, offers automatic differentiation and reusable layer blocks, and exposes C++, BrainScript, C# and Python APIs. Microsoft reports a surprise result of <4% ImageNet error and ~6% speech transcription error using the toolkit. The demoed system is a real‑time video pipeline for helmet detection, abandoned‑object and vehicle/human classification built in a few months on GPU clusters, and models run for inference on CPUs.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:08Microsoft cognitive toolkit
- 5:57Framework capabilities
- 10:30Model architecture examples
- 13:36Computer vision demos
- 20:27Resources and deployment
- 24:15Q&A session
related talks
01h 57mUsing CNTKs Python Interface for Deep Learning
55mDeep Learning with CNTK and F#
58mDeep Learning in Microsoft Azure: CNTK, CaffeOnSpark and Tensorflow
33mBuilding end-to-end computer vision solutions from deep learning models
57mDeep Learning with CNTK and F#
55mA Tour of Deep Learning With C++
57mDeep Learning with CNTK and F#
34mFaster deep learning solutions...