Ensuring Quality Data for Deep Learning in Varied Application Domains: Data Collection, Curation and Annotation


Date:  Tuesday, May 17 (Main Conference Day 1)

Start Time: 4:50 pm

End Time: 5:55 pm

In this presentation, we explore the data lifecycle for deep learning with special emphasis on data curation and how to ensure quality annotations. For improving data curation, we will examine techniques like active learning and focus on how we choose which data to send for annotation. We’ll also discuss how to select an annotation partner and how to efficiently do annotation in-house. And we will detail how to frame good annotation instructions for different annotation tasks like lidar annotation, semantic segmentation and sequence annotations. In addition, we explain common problems seen in the curation and annotation process and how to overcome them.

Track

Session Speakers

  • Gaurav Singh is an engineering leader for Nemo product at Ridecell which aims to make automotive data more meaningful and searchable. He works with his team to solve key problems related to architecture, perception and systems for both the edge and cloud parts of Nemo. He is especially interested in the drive to improve edge processing in vehicles to increase safety and to identify useful data to send back to the cloud. Gaurav has seven years of experience in developing and deploying computer vision-based perception systems for ADAS and AVs. Through his early career experience in developing system software, Gaurav brings an extensive systems background to his current job in the computer vision field. He has a Masters degree in Robotics from Carnegie Mellon University.


Share

Ensuring Quality Data for Deep Learning in Varied Application Domains: Data Collection, Curation and Annotation