l3p-cv/lost
Label Objects and Save Time (LOST) - Design your own smart Image Annotation process in a web-based environment.
What it solves
LOST simplifies the process of image annotation for machine learning datasets. It addresses the complexity of managing collaborative labeling efforts and the need for semi-automatic workflows to reduce the manual effort required to label images.
How it works
It is a web-based framework that allows users to create annotation pipelines. These pipelines can be either "out of the box" for immediate use or customized by combining different tools and algorithms. The system supports semi-automatic annotation by presenting AI-generated proposals to human annotators for verification or correction.
Who it’s for
It is designed for researchers and developers who need to label image data, specifically those working with bounding boxes, polygons, points, or lines, and those who need to distribute labeling tasks across a global team of annotators.
Highlights
- Collaborative Workflow: Distribute annotation tasks to users worldwide via a web browser.
- Flexible Pipelines: Support for both predefined and user-defined annotation processes.
- Cloud Integration: Direct connectivity to external storage like AWS S3, Azure Blobstorage, and FTP servers.
- Semi-Automatic Labeling: Ability to integrate AI proposals to speed up the manual labeling process.
- Scalable Design: Intensive computing processes can be distributed across multiple machines.