Synthmark is a production-grade annotation workspace designed specifically for computer vision teams. It streamlines the process of labeling images with polygon, bounding box, and keypoint annotations while providing built-in review workflows, version control, and ML-ready exports.
Building a high-performance canvas-based annotation tool that handles large datasets while maintaining smooth interactions across polygon drawing, box selection, and keypoint placement. The challenge was implementing real-time collaboration features, role-based access control, and ensuring exports are compatible with major ML frameworks.
Delivered a full-stack SaaS platform with a custom annotation canvas, email-based verification flow, admin dashboard for user management, and support for COCO, YOLO, VOC, and JSON exports. The platform features role-based access (admin, annotators, reviewers) and intuitive keyboard shortcuts.
Precise multi-point polygon drawing with vertex editing
Fast box creation with aspect ratio locking
Skeleton-based point labeling for poses and landmarks
Built-in review cycle with changeset tracking
Admin, Annotator, and Reviewer roles with granular permissions
COCO, YOLO, VOC, and JSON format support
Efficient labeling with customizable shortcuts
User management and project oversight
Secure signup flow with verification codes
Track annotation history and changes
