Prexable Successfully Concludes Phase III with Launch of Advanced Video Annotation Platform
Prexable Successfully Concludes Phase III with Launch of Advanced Video Annotation Platform
Prexable is excited to announce the successful completion of its ambitious Phase 3 development cycle, which commenced in early 2025. This phase delivered a high-impact, user-facing feature critical for the machine learning community and included a major strategic architectural upgrade.
Management Review and Technical Session
On Sunday, October 5, 2025, an online executive meeting was held via Google Meet to review the outcomes and operational readiness of Phase III. The session included:
- Boshra Rajaei (CEO)
- Zahra Rajaei (CTO)
- Sadegh Ghafari (CBO)
- Hamid Golyani (CFO)
The primary focus was an in-depth review and brainstorming session for the new Video Annotation feature, followed by a discussion on the successful frontend refactoring.
Phase III Key Highlights
1. Launch of the Video Annotation Feature
The centerpiece of Phase III is the introduction of a powerful Video Annotation Feature. This tool is specifically designed to streamline the creation of high-quality training datasets for complex Computer Vision models.
Core Annotation Workflow:
- Users can upload videos and define their desired frame-rate for processing and sampling.
- The annotation process leverages temporal coherence: users first annotate objects on the initial frame.
- For subsequent frames, users can efficiently propagate the existing annotations, easily correcting them (e.g., for moving or deforming objects) with simple mouse gestures, or creating new annotations as needed. This significantly reduces manual effort for sequential frames.
- Upon completion, users can download the fully annotated video file and the corresponding annotation file separately.
- Users also can edit their annotations anytime later.
Technical Capabilities and Machine Learning Integration:
- Supported Primitives: The tool provides flexibility with three essential geometric shapes: Rectangles (for Bounding Box detection), Polygons (for Instance Segmentation) and Circles.
- COCO Compatibility: All exported JSON annotation files are generated to be fully compliant with the COCO (Common Objects in Context) format. This ensures seamless integration, allowing users to immediately utilize the datasets to train, evaluate, and fine-tune their custom object detection and segmentation models.
The session included an in-depth brainstorming session to detail further enhancements and potential capabilities for the annotation pipeline.
2. Frontend Architectural Refactoring to Strapi
To ensure scalability and future agility, a significant structural overhaul was completed. The original WordPress frontend was entirely refactored and successfully migrated to Strapi, a modern headless Content Management System (CMS). This transition establishes a robust, decoupled architecture, enabling Prexable to rapidly deploy future feature iterations.
With Phase III successfully deployed, Prexable is now dedicating its full resources to planning and executing Phase IV.