Computer Vision Workflows Workshop with FiftyOne
A free, virtual workshop on April 8th to learn how to use the open-source FiftyOne app for curating, visualizing, and evaluating computer vision models.
Why it matters
This workshop provides valuable hands-on training for anyone working on computer vision projects, helping them improve data management, model development, and evaluation workflows.
Key Points
- 1Structure unstructured data into a queryable schema for images, videos, and point clouds
- 2Use the FiftyOne SDK to create complex data views and filter based on model predictions, labels, and confidence scores
- 3Visualize high-dimensional data embeddings to identify data gaps and outliers using FiftyOne Brain
- 4Automate data curation by selecting diverse subsets and prioritizing high-entropy samples to reduce labeling costs
Details
This workshop will provide hands-on training on the FiftyOne open-source app, which helps data scientists and ML engineers manage the entire computer vision model development lifecycle. Attendees will learn how to structure unstructured data into a queryable schema, use the FiftyOne SDK to create custom data views, visualize high-dimensional embeddings to understand dataset characteristics, and implement automated data curation techniques. The workshop will also cover using FiftyOne to debug model performance by generating evaluation metrics and visualizing model failures directly in the app. Customizing FiftyOne for domain-specific tasks will also be discussed.
No comments yet
Be the first to comment