Timothy Norton

Key Skills

Software

CVAT

Data Annotation Tech

Labelbox

Scale AI

Surge AI

Other

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Medical Dicom

Video

Top Label Types

Bounding Box

Segmentation

Object Detection

Tracking

Freelancer Overview

I am a freelance AI training specialist with over five years of hands-on experience in data labeling, annotation, and dataset validation for computer vision and NLP applications. My expertise spans high-volume image, video, and audio annotation—including bounding boxes, polygon segmentation, multi-object tracking, and timestamped transcription—using industry-standard platforms like Labelbox, CVAT, Supervisely, and Amazon SageMaker Ground Truth. I have delivered large-scale, YOLO-ready datasets for domains such as autonomous driving, retail detection, and surveillance, consistently maintaining 98–99% annotation accuracy and meeting rigorous quality standards. My background in software engineering allows me to automate data pipelines, implement robust QA checks, and optimize annotation workflows. Notable projects include leading 120k+ bounding box annotations for vehicle detection and multi-label emotion tagging for speech datasets, resulting in measurable improvements to model performance. I am highly skilled at interpreting complex guidelines, identifying edge cases, and ensuring data quality throughout the AI training lifecycle.

Labeling Experience

Multi-Object Video Tracking & ID Persistence (Autonomous Driving)

CVATVideoBounding BoxObject Detection

Performed frame-by-frame multi-object tracking across urban driving video sequences. Assigned persistent IDs to vehicles, pedestrians, and cyclists across long clips with occlusions, motion blur, and re-entry events. Handled complex edge cases including partial visibility, overlapping objects, and direction changes. Ensured temporal consistency and reduced ID-switch errors to improve downstream Deep SORT integration. Contributed to YOLOv8-based detection + tracking pipeline, supporting 0.91 mAP performance. Maintained 98%+ QA accuracy under strict SLA requirements.

Performed frame-by-frame multi-object tracking across urban driving video sequences. Assigned persistent IDs to vehicles, pedestrians, and cyclists across long clips with occlusions, motion blur, and re-entry events. Handled complex edge cases including partial visibility, overlapping objects, and direction changes. Ensured temporal consistency and reduced ID-switch errors to improve downstream Deep SORT integration. Contributed to YOLOv8-based detection + tracking pipeline, supporting 0.91 mAP performance. Maintained 98%+ QA accuracy under strict SLA requirements.

2022 - 2024

Autonomous Vehicle Multi-Object Detection & Tracking Dataset (YOLOv8 Optimized)

CVATImageBounding BoxSegmentation

Annotated and validated 120,000+ bounding boxes for autonomous driving datasets. Performed multi-class object detection (vehicles, pedestrians, traffic signals) and frame-by-frame video tracking with ID consistency. Delivered YOLOv8-ready datasets (normalized coordinates, class mapping) and supported mAP evaluation (0.91 achieved). Applied IoU-based QA checks, maintained 98–99% accuracy, and met strict SLA timelines.

Annotated and validated 120,000+ bounding boxes for autonomous driving datasets. Performed multi-class object detection (vehicles, pedestrians, traffic signals) and frame-by-frame video tracking with ID consistency. Delivered YOLOv8-ready datasets (normalized coordinates, class mapping) and supported mAP evaluation (0.91 achieved). Applied IoU-based QA checks, maintained 98–99% accuracy, and met strict SLA timelines.

2022 - 2023

Education

Work History