Tag and detect: Object detection made easy
In recent years, machine learning-based algorithms have become a popular choice for solving computer vision tasks. Training these algorithms, however, often requires a large quantity of labeled data that is traditionally difficult to prepare. In this talk, we present a recent collaboration between the Microsoft Partner Catalyst team and InSoundz, an audio-tracking startup, to build an Object Detection System using CNTK, Microsoft’s Deep Learning framework. As part of the project, the Partner Catalyst Team has developed the “Visual Object Tagging Tool,” which provides an end-to-end solution for building training sets for Object Detection algorithms from video files, as well as the ability to review the Object Detection Model on new data. As part of the talk, we demonstrate how to build an Object Detection Model using CNTK and the Visual Object Tagging Tool.