Ziqi Li

Key Skills

Software

Roboflow

Top Subject Matter

No subject matter listed

Top Data Types

Image

Video

Top Label Types

Bounding Box

Polygon

Classification

Tracking

Data Collection

Freelancer Overview

I have a strong background in data annotation and AI training data, particularly within computer vision and medical domains. My experience includes developing and implementing data collection and annotation workflows for deep learning projects, such as training YOLO-based object detectors for robotic manipulation and segmenting medical images using CNNs and U-Net. I have hands-on expertise in designing targeted labeling methods, managing large-scale datasets, and applying data augmentation and transfer learning to improve model robustness. My technical skills include Python, SQL, Docker, TensorFlow, PyTorch, and Scikit-learn, and I am comfortable working in Linux environments. I am passionate about ensuring data quality and reproducibility, and have contributed to research projects that resulted in high-precision models and published papers in top journals.

Labeling Experience

Data Annotation Experience in Mobile Manipulators for Nursing Assistance

RoboflowVideoBounding BoxPolygon

This project encompassed the entire pipeline from data collection to deployment, focusing on gathering images of tableware, annotating them for object detection, training a YOLO-based detector integrated with a ROS2 planner, and deploying the system using Docker for reproducibility. Data labeling involved identifying various tableware items, drawing bounding boxes around them, and conducting a quality assurance review to ensure accuracy and consistency in the annotations. The project included the collection of over 5,000 images and approximately 20 hours dedicated to labeling and reviews, with collaboration from a team of four individuals to enrich the process. Quality measures included establishing clear annotation guidelines, targeting a 90% inter-annotator agreement, implementing a two-tier review process for accuracy, and conducting integrity checks to verify completeness and correctness of the annotations.

This project encompassed the entire pipeline from data collection to deployment, focusing on gathering images of tableware, annotating them for object detection, training a YOLO-based detector integrated with a ROS2 planner, and deploying the system using Docker for reproducibility. Data labeling involved identifying various tableware items, drawing bounding boxes around them, and conducting a quality assurance review to ensure accuracy and consistency in the annotations. The project included the collection of over 5,000 images and approximately 20 hours dedicated to labeling and reviews, with collaboration from a team of four individuals to enrich the process. Quality measures included establishing clear annotation guidelines, targeting a 90% inter-annotator agreement, implementing a two-tier review process for accuracy, and conducting integrity checks to verify completeness and correctness of the annotations.

2023 - 2024

Education

Work History