Zhen Z.

Key Skills

Software

Other

OpenCV AI Kit (OAK)

Top Subject Matter

Autonomous driving / LiDAR-based navigation (simulation)

Machine learning for vibroacoustic signal analysis

Image and video compression quality optimization

Top Data Types

3D Sensor

Audio

Image

Top Task Types

Classification

Object Detection

Freelancer Overview

Master Thesis (TUM) — Spiking Neural Network for Autonomous Navigation based on LiDAR Sensor (R-STDP + CoppeliaSim/Carla. Brings 10+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other and OpenCV AI Kit (OAK). Education includes Master of Science, Technische Universität München (TUM) (2022) and Bachelor of Science, Wuhan University of Technology (211) (2018). AI-training focus includes data types such as 3D Sensor, Audio, and Image and labeling workflows including Computer Programming, Coding, and Classification.

Labeling Experience

Master Thesis (TUM) — Spiking Neural Network for Autonomous Navigation based on LiDAR Sensor (R-STDP + CoppeliaSim/Carla lane following)

Other3D Sensor

Conducted AI/ML experimentation for autonomous navigation using LiDAR sensor inputs as part of a research master thesis. Implemented a Spiking Neural Network approach combining R-STDP with simulation environments to support lane-following behavior. The work required preparing and running simulation-based datasets and evaluating model performance for navigation tasks. • Data preparation and preprocessing for LiDAR-based inputs • Simulation-driven testing using CoppeliaSim/Carla • Modeling and training/inference logic for R-STDP SNN • Performance evaluation for lane following and navigation results

2022 - 2022

Research Practice Student (Machine Learning) - Technische Universität München (TUM)

ImageClassification

As a Research Practice student at TUM, you conducted unsupervised machine learning research on vibroacoustic signal classification. You handled the end-to-end preprocessing of data and applied representation learning with an autoencoder. You then used clustering methods to discover structure in the signals, requiring solid ML fundamentals and experimental rigor. • Preprocessed vibroacoustic datasets for unsupervised learning • Implemented an AutoEncoder for feature learning • Applied clustering algorithms for categorization • Analyzed results and iterated on modeling choices

2021 - 2022

Research Practice (TUM) — Unsupervised Classification of Vibroacoustic Signals based on Machine Learning

OtherAudioClassification

Performed unsupervised machine learning on vibroacoustic signal data using deep learning components such as autoencoders and clustering. Focused on preprocessing and transforming raw signals into features suitable for model training and clustering. The research involved running experiments to group similar signal patterns without explicit labels. • Data preprocessing for vibroacoustic signals • AutoEncoder-based feature learning • Clustering algorithm for unsupervised categorization • Experiment iterations and analysis of clustering outcomes

2021 - 2022

Practical Training (TUM) — Image and Video Compression Lab (PSNR optimization using JPEG quantization tables)

OpenCV AI Kit (OAK)Image

Worked in an image and video compression laboratory to optimize objective image-quality metrics for still images and videos. Implemented and tuned JPEG quantization-table strategies and assessed resulting quality using PSNR as the evaluation metric. The role involved preparing image/video inputs and running compression experiments to measure improvements. • Optimization of PSNR via JPEG quantization tables • Image/video compression experimentation • Using ML/optimization workflows for quality improvement • Evaluation of compression quality across test inputs

2020 - 2020

Practical Training (TUM) — Opponent Detection and Localization (NAO Robot-ROS) / Humanoid RoboCup

Other3D SensorObject Detection

Developed and tested CNN and OpenCV-based methods for opponent detection and self-localization in humanoid RoboCup contexts. Worked with a NAO robot integrated with ROS, which required running perception pipelines over sensor data streams. The experiments involved preparing inputs for detection/localization and evaluating model outputs in robot scenarios. • CNN/OpenCV-based opponent detection pipeline • ROS/Ubuntu-based deployment on NAO robot • Self-localization for humanoid RoboCup environment • Experimental evaluation of detection and localization results

2019 - 2020

Education

Work History