Jude

Key Skills

Software

No software listed

Top Subject Matter

NLP sentiment/emotion classification

Routing/path planning optimization for inspection

Top Data Types

Text

Top Task Types

Emotion Recognition

Freelancer Overview

中文多标签情绪检测（课程设计）. Brings 6+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include PyTorch and MATLAB. Education includes Bachelor of Science, China Civil Aviation University (2024). AI-training focus includes data types such as Text, Computer Code, and Programming and labeling workflows including Emotion Recognition, Computer Programming, and Coding.

Labeling Experience

中文多标签情绪检测（课程设计）

TextEmotion Recognition

The project built a Chinese multi-label emotion recognition classifier using RoBERTa and addressed label imbalance and varying label-trigger confidence. It enhanced training data via Chinese-English back-translation and used loss weighting plus FocalLoss to improve recognition of rare emotion categories. The model performed dynamic per-label threshold searching during inference to better match each label’s optimal decision boundary. • Data augmentation using Chinese-English back-translation • Tail-category improvement via class-frequency weights and FocalLoss • Dynamic threshold search per emotion label at prediction time • Evaluation on SemEval official dataset using F1-macro

2024 - 2025

狭窄空间内飞机蒙皮检测无人机覆盖规划研究（毕业设计）

The work focused on iterative UAV path planning optimization for full coverage and collision-free traversal of viewpoints around an aircraft skin inspection scenario. While it did not directly describe data labeling, it involved constructing cost matrices, running LKH for global route generation, and using RRT for local collision-avoidance path generation. The resulting optimized traversal paths improved coverage efficiency over baseline approaches. • Built initial shortest paths using cost matrices and LKH • Generated local obstacle-avoidance paths with RRT and collision checking • Iteratively updated cost matrices to refine routing • Achieved improved traversal performance across 165 viewpoints

2023 - 2024

Education

Work History