Gökçe Keleşyılmaz

Key Skills

Software

Roboflow

CVAT

Top Subject Matter

Computer Vision

Sports Analytics

Video-Image Analysis

Top Data Types

Video

Image

Computer Code Programming

Top Task Types

Object Detection

Bounding Box

Segmentation

Classification

Data Collection

Computer Programming/Coding

Freelancer Overview

I have over 2 years of hands-on experience in data labeling and AI training workflows, with a strong focus on video annotation and object detection tasks. I contributed to a Basketball Video Analysis Project, where I performed precise frame-by-frame labeling and ensured high-quality annotations for model training. I am experienced with tools such as Roboflow and follow structured, detail-oriented workflows to maintain consistency and accuracy. I hold a Bachelor’s degree (2025) and am currently pursuing a Master’s degree (2026) at İnönü University. My background combines academic knowledge with practical AI training data experience, particularly in video-based datasets. I am reliable, efficient, and committed to delivering high-quality labeled data that supports robust machine learning models.

Labeling Experience

Turkish Sign Language Gestures Project Data labeling and YOLO (İşaretSensin)

ImageObject Detection

İşaretSensin is an accessibility-focused computer vision project that leverages a YOLOv8-based model to detect and recognize Turkish Sign Language gestures in real time. As part of the project, I created a custom dataset by capturing original images and annotating them using Roboflow, ensuring high-quality and consistent labeling. To improve model performance and generalization, I combined my custom dataset with an existing public dataset and trained a YOLO-based object detection model. This end-to-end process involved data collection, annotation, dataset optimization, and model training, demonstrating strong practical experience in building and scaling AI training data pipelines.

İşaretSensin is an accessibility-focused computer vision project that leverages a YOLOv8-based model to detect and recognize Turkish Sign Language gestures in real time. As part of the project, I created a custom dataset by capturing original images and annotating them using Roboflow, ensuring high-quality and consistent labeling. To improve model performance and generalization, I combined my custom dataset with an existing public dataset and trained a YOLO-based object detection model. This end-to-end process involved data collection, annotation, dataset optimization, and model training, demonstrating strong practical experience in building and scaling AI training data pipelines.

2025 - 2025

Basketball Video Analysis Project - Video Annotation & Object Detection Labeler

RoboflowVideoObject Detection

As part of a basketball video analysis project, I collected and annotated basketball game videos to create labeled datasets for training a custom YOLO model. The task involved manually labeling key objects such as the ball and rim within video frames to enable downstream object detection and tracking tasks. Using Roboflow as the annotation tool, I ensured high-quality, consistent labels for robust model training. • Labeled and annotated objects in basketball video frames using bounding boxes. • Utilized Roboflow as the primary annotation and dataset management platform. • Prepared data for training YOLO-based object detection and ball tracking models. • Ensured annotation consistency and accuracy across all labeled samples.

As part of a basketball video analysis project, I collected and annotated basketball game videos to create labeled datasets for training a custom YOLO model. The task involved manually labeling key objects such as the ball and rim within video frames to enable downstream object detection and tracking tasks. Using Roboflow as the annotation tool, I ensured high-quality, consistent labels for robust model training. • Labeled and annotated objects in basketball video frames using bounding boxes. • Utilized Roboflow as the primary annotation and dataset management platform. • Prepared data for training YOLO-based object detection and ball tracking models. • Ensured annotation consistency and accuracy across all labeled samples.

Not specified

Education

Work History