Sahr Fomba Gindeh

Key Skills

Software

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Video

Top Label Types

Bounding Box

Polygon

Object Detection

Tracking

Freelancer Overview

I am a detail-oriented data annotator and AI training data specialist with hands-on experience in image and video annotation, quality assurance, and AI output evaluation. My work involves using tools like CVAT, LabelImg, Google Sheets, and Excel to deliver high-accuracy annotations with over 98% guideline adherence, supporting supervised fine-tuning and prompt evaluation tasks. I have developed a portfolio of computer vision projects, applying bounding boxes, polygons, and keypoints to diverse datasets while maintaining strict QA standards. My strengths include critical thinking, high-precision writing, and the ability to work efficiently in remote, asynchronous environments. I am committed to producing reliable, high-quality training data to support machine learning workflows.

Labeling Experience

Traffic & Object Detection for Computer Vision

CVATVideoBounding BoxPolygon

* Executed high-precision video annotation tasks using CVAT for object detection and tracking models. Specialized in frame-by-frame analysis to identify vehicles, pedestrians, and cyclists in complex urban environments. Key Tasks: Dynamic Bounding Boxes: Annotated moving objects with strict adherence to occlusion rules and tight-fitting borders. Polygon Segmentation: Created pixel-perfect masks for irregular shapes to support semantic segmentation training. Quality Control: Maintained a 98% accuracy rate, self-auditing for jitter and drift in temporal tracking. Workflow: Managed data pipelines via Google Sheets to track progress and flag edge cases.

* Executed high-precision video annotation tasks using CVAT for object detection and tracking models. Specialized in frame-by-frame analysis to identify vehicles, pedestrians, and cyclists in complex urban environments. Key Tasks: Dynamic Bounding Boxes: Annotated moving objects with strict adherence to occlusion rules and tight-fitting borders. Polygon Segmentation: Created pixel-perfect masks for irregular shapes to support semantic segmentation training. Quality Control: Maintained a 98% accuracy rate, self-auditing for jitter and drift in temporal tracking. Workflow: Managed data pipelines via Google Sheets to track progress and flag edge cases.

2024 - 2025

Education

Work History