ISE Developer Blog

Computer Vision - ISE Developer Blog

Detecting “Action” and “Cut” in Archival Footage Using a Multi-model Computer Vision and Audio Approach with Azure Cognitive Services

September 27, 2021 Sep 27, 2021 09/27/21
Nile Wilson
Movies and TV shows require multiple takes per scene and may have a substantial amount of archival footage as a result. Here, we use Azure Cognitive Services and custom code to develop a multi-model Machine Learning (ML) solution to automatically detect discardable footage to save media companies manual archiving hours and storage space.

Building an Action Detection Scoring Pipeline for Digital Dailies

September 20, 2021 Sep 20, 2021 09/20/21
Samuel Mendenhall
Media companies capture footage filmed for the entire day in what's known as ‘digital dailies’. When talking about terabytes and petabytes of content, storage costs can be a factor. Lets explore Machine Learning approaches to identify which content can be archived or discarded which will save on those storage costs.

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.