Qing G.

Key Skills

Software

Don't disclose

CVAT

Top Subject Matter

Crime-scene investigation records and large-model training data

Floor-plan detection

room segmentation

Top Data Types

Text

Image

Computer Code Programming

Top Task Types

Prompt + Response Writing (SFT)

Segmentation

Bounding Box

Polygon

Classification

Point/Key Point

Fine-tuning

Computer Programming/Coding

Object Detection

Text Generation

Freelancer Overview

Crime-Scene Investigation Records for Public Security Cases (AI dataset generation, cleaning, quantization fine-tuning, . Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Don't disclose. Education includes Master's Degree, Hunan University of Science and Technology (2020). AI-training focus includes data types such as Text and Image and labeling workflows including Prompt + Response Writing (SFT) and Segmentation.

Labeling Experience

Cultural Relic Line Drawings and Pose-Based 3D Spatial Modeling (dataset generation and model training for line-drawing masks)

Don't discloseImageSegmentation

Created line-drawing mask training data using a large model with image+text to image generation. Trained an image-to-image model to generate line-drawing masks from input images. Combined the generated line drawings with depth maps and poses to complete 3D spatial modeling output. • Generated a line-drawing mask dataset via large-model generation. • Trained an image-to-image model for producing line-drawing masks. • Fused line drawings, depth maps, and pose information for 3D modeling. • Produced usable structured annotations for downstream 3D reconstruction.

2026 - 2026

Floor-Plan Detection Based on Point-Cloud Projection (mask generation, segmentation refinement, and spatial annotation tasks)

Don't discloseImageSegmentation

Performed floor-plan and room layout analysis by generating and refining segmentation masks for room and block regions. Used a large model to separate connected regions and normalize geometric shapes before classifying blocks and objects. Refined wall lines based on the segmentation and spatial modeling outputs. • Segmented mask blocks accurately for room/area representation. • Applied large-model-assisted separation of connected blocks and normalization of shapes. • Classified blocks and segmented furniture, including door recognition. • Refined wall lines from processed spatial geometry.

2025 - 2026

Crime-Scene Investigation Records for Public Security Cases (AI dataset generation, cleaning, quantization fine-tuning, and deployment)

Don't discloseTextPrompt + Response Writing (SFT)

Built a crime-scene investigation record dataset using a large model with image+text to text generation. Cleaned, corrected, and prepared the generated dataset for downstream training and usage. Fine-tuned a quantized model and validated it via deployment and API testing. • Generated structured case documentation from images and prompts. • Performed dataset cleaning and correction to improve label/data quality. • Fine-tuned a quantized model based on the prepared dataset. • Deployed with Docker, tested with Dify, and exposed functionality via HTTP invocation.

2025 - 2026

Education

Work History