Overview
Explore the critical role of dataset management in delivering successful computer vision solutions in this 41-minute conference talk. Learn how to disentangle business goals from technical implementations, express assumptions through dataset construction, and control for corner cases. Discover strategies for updating datasets to reflect new insights and evolving data. Gain valuable insights on balancing the emphasis between modeling and dataset management, and understand why a well-maintained dataset is often the key to project success. Delve into real-life examples from projects implemented for large organizations, and acquire practical knowledge on feature engineering, addressing the gap between academic research and real-world applications, and managing data drift in computer vision projects.
Syllabus
Introduction
Agenda
What is Computer Vision
Deep Learning vs Machine Learning
Feature Engineering
Academia vs Real World
Key Point
Use Case
How to Action
Example
Data Drift
Taught by
Open Data Science