Deep Learning

Deploying a Batch AI Cluster for Distributed Deep Learning Model Training

July 9, 2018 Jul 9, 2018 07/9/18

Stephanie Marker

Microsoft and Land O'Lakes partnered to develop an automated solution to identify sustainable farming practices given thousands of satellite images of Iowan farms. Our primary goal was to reduce the reliance on manual interviewing of farmers and make it more profitable for farmers to follow sustainable farming practices. To tackle this issue our team deployed a highly scalable Batch AI cluster on Azure and then performed distributed deep learning model training with Horovod.

Using Otsu’s method to generate data for training of deep learning image segmentation models

May 17, 2018 May 17, 2018 05/17/18

Clemens Wolff

In this article, we introduce a technique to rapidly pre-label training data for image segmentation models such that annotators no longer have to painstakingly hand-annotate every pixel of interest in an image. The approach is implemented in Python and OpenCV and extensible to any image segmentation task that aims to identify a subset of visually distinct pixels in an image.

Fashion Retail Inventory Management With Deep Learning Content-based Image Retrieval

May 15, 2018 May 15, 2018 05/15/18

CY Yam

In early 2018, Microsoft partnered with a successful international online fashion retailer to develop an efficient system for logging newly arrived apparel items and quickly determining whether the item is already in stock

Deep Learning Image Segmentation for Ecommerce Catalogue Visual Search

April 18, 2018 Apr 18, 2018 04/18/18

CY Yam

We explored possible techniques to identify duplicate catalogue entries in fashion retail through image similarity with deep learning and computer vision.

Sequence Intent Classification Using Hierarchical Attention Networks

March 6, 2018 Mar 6, 2018 03/6/18

Olga Liakhovich

We analyze how Hierarchical Attention Neural Networks could be helpful with malware detection and classification scenarios, demonstrating the usefulness of this approach for generic sequence intent analysis. The novelty of our approach is in applying techniques that are used to discover structure in a narrative text to data that describes the behavior of executables.

Comparing Image-Classification Systems: Custom Vision Service vs. Inception

December 5, 2017 Dec 5, 2017 12/5/17

Clemens Wolff

This story covers how to get started with transfer-learning and build image classification models in Python with the Custom Vision Service. We compare the results with the popular Tensorflow-based models Inception and MobileNet.

Stock Market Predictions with Natural Language Deep Learning

December 4, 2017 Dec 4, 2017 12/4/17

Patty Ryan

We developed a deep learning model using a one-dimensional convolutional neural network to predict future stock market performance of companies using Azure Machine Learning Workbench and Keras.

Autoscaling Deep Learning Training with Kubernetes

November 21, 2017 Nov 21, 2017 11/21/17

Rita Zhang

We explore how we worked with a customer to add autoscaling capability to a Kubernetes cluster to meet bursty demands for deep learning training in a cost-efficient manner.

Learning Image to Image Translation with Generative Adversarial Networks

June 12, 2017 Jun 12, 2017 06/12/17

Olga Liakhovich

Microsoft has partnered with Getty Images to explore how Neural Nets could be used to transform the stock photo industry.

Classifying Leaks Using Cognitive Toolkit

May 10, 2017 May 10, 2017 05/10/17

Michael Lanzetta

We use Deep Learning to turn a painful and time-consuming leak-detection task for water and oil pipelines into a fast, painless process. Using Python and Fast Fourier Transforms, we turn audio sensor data into images, then use Convolutional Neural Networks to detect and classify pipeline anomalies.

CSE Developer Blog

Deep Learning - CSE Developer Blog