Deep Learning Based Object Detection (Fast R-CNN) in the Microsoft Cognitive Toolkit

  • R1
  • Day 2, 17:25‑18:10
  • Chinese talk w. English slides
  • Data Analysis
  • Intermediate

This session showcases work on generic object detection using deep learning. We use the state-of-the-art object detector in the Microsoft Cognitive Toolkit (CNTK) called Fast R-CNN based on Python. Fast R-CNN is based on a Deep Neural Network which was pre-trained on millions of images.

Talk Detail

The objective of the talk is to describe the DNN technology and show the full object detection pipeline encompassing image annotation, model training, evaluation. Fast R-CNN is a fast framework for object detection with deep ConvNets.Fast R-CNN was initially described in an [arXiv tech report](http://arxiv.org/abs/1504.08083) and later published at ICCV 2015. Microsoft Cognitive Toolkit (CNTK), an MIT Licensed open source deep-learning toolkitthat describes neural networks as a series of computational steps via a directed graph. It supports both Linux and Windows. Based on researchers at HKBU, CNTK’s LSTM performance is 5-10x faster than the other toolkits. The session introduct how to combine the state of art algoritm and toolkit to do image Object Detection.

Slides Link

Speaker Information

Herman Wu

Been a Java Developers for 3 years.
Work as a Microsoft consults for 5 years and work as a Technical Evangelist for the other 5 years.