Research Experiences


------------ Research Service & Publications ------------

ROS-based Formation Control and Collaborative Tracking

Undergraduate Final-year Project (2020.12 -- 2021/5)

❤ Self-built three original Gazebo scenarios including Playground, Gas Station, and House. Explored 5 diverse intelligent vehicles via ROS systems.
✿ Implemented multiple algorthms: (1) Multi-agent PID formation control and collaborative tracking. (2) Visual tracking and ROS node tracking. (3) YOLO-V3 object detection controlled mission and multi-vehicle systems with autonomous navigation.

Action Recognition for Skateboarding & Video BGM Music Generation

Research Internship at the Shanghai Qizhi Institute. (2021/7 -- 2021/12)

❤ Motivated by special love for videos and skateboarding, I self-curated the custom "SkateboardAI" datasets and implemented multiple uni-modal and multi-modal methods for trick recognition.
✿ Video BGM music generation: Successfully explored several music models (e.g., CMT, REMI, Transformers) to design new pipelines.

Amazing Robotic Arm with Human-robot Interaction Applications

[1] Summer Internship at CUHK Shenzhen Research Institute. (2022/5 -- 2022/9)
[2] Master Summer Internship at Shanghai Jiao Tong University. (2023/5 -- 2023/7)

❤ Master the usage of Franka Emika Panda & Kinova Gen3 Lite & Kinova Gen3. Explore amazing robotic applications such as grasping, writing, cooking, or conducting complicated human-robot interaction tasks.
✿ Aimed to connect robots with human for brilliant interactions with ChatGPT LLM.

Cambridge AI Program

Independent Research.
(2020/6 -- 2020/9)

❤ Explored multiple robotic manipulation tasks like grasping, button-pushing and door-opening with reinforcement learning, state representation learning, and imitation learning.
✿ Self-conducted whole experiments and complete the academic paper, which was accepted by the AAAI 2021 Student Abstract and Poster Program. Achieved the Cambridge CCISTC AI Program Certification. (Excellent research assistant.)

Car Damage Detection

The Official UCInspire Summer Research Program [2020 Online]
(2020/7 -- 2020/10)

❤ Built a custom Mask R-CNN Model to detect car damage areas.
✿ Propose a new image alignment method inspired by self-supervised learning to find perspective transformation between pre/post car images.

Community Detection & Big Data

Group Research. (2018/10 -- 2019/1)
Chinese Patent.

❤ Researched on Attributed Community Search (ACS) problem in Big Data field, which aims to help users find a personalized community containing the query nodes with high relevance of query attributes in convoluted real-world networks.
✿ Proposed an effective Attributed Community Search Algorithm --- RACSF-ACS, responsible for part of algorithm design and scientific writing.

Semi-supervised learning & Computer Vision

Independent Research. (2019/9 -- 2020/3) First-author's paper.

❤ Driven by pure passion for computer vision, I dived into interesting image classification, object detection, video segmentation.
✿ This work focuses on semi-supervised learning with self-supervised proxy tasks ( i.e. image colorization, image rotation, geometric transformation). Fully responsible for the algorithm design, research experiments and academic writing.

Automatic Machine Learning & HPO

Group Research. (2019/3 -- 2019/7)
Information Sciences (SCI), Chinese Patent.

❤ Focused on Hyperparameter Optimization with budget constraints and proposed the algorithm "ExperienceThinking" and relevant definitions.
✿ Built two intelligent modules "Human Experience" and "Parameter Analysis" which serve as our left and right brains to effectively infer optimal hyperparameter configurations for machine learning problems.

Engineering Projects



Intriguing engineering projects inspire me to explore more knowledge in various fields like Robotics, Self-supervised representation learning in Computer Vision, Reinforcement Learning, and Self-driving Car.

Gesture Tracking Robotic Arm

"Provincial Science and Technology Innovation Project". (2018.11 -- 2019.11) [Project Leader]
☘ Built an intelligent robotic arm on Arduino platform with Hitec Series servos and Leap Motion Controller which can utilize humans' gestures to control its posture in Intelligent Control Laboratory.
Fully Responsible for managing the project process and completing software part on Arduino Platform and Leap Motion. ☕ Project Code.
Tools: Arduino, Javascript, Node.js, Leap Motion.

Image Colorization in Computer Vision.

Personal Project on Computer Vision.
☘ Attracted by the amazing paper Tracking Emerges by Colorizing Videos, I also trained the Encoder-Decoder Networks to colorize the gray pictures in the Lab Space. Then utilized such novel tasks in my independent research on semi-supervised image classification.
Image colorization seems flexible and provides us multiple results for one picture. ☕ Project Code.
Tools: Python, Tensorflow, Keras, OpenCV.

Object Detection on Open Images 2019

Independent Project in Computer Vision. [Kaggle Competition]
☘ Taken part in the kaggle project "Open Images Track 2019 -- Object Detection" which is organized by Google Research. Attempted to solve the Object detection track challenge , utilized Tensorflow Hub Module of pre-trained Faster-RCNN and SSD-MobileNet models to detect objects in varied and complex images in Open Images.☕ Project Code.
Tools: Python, Tensorflow Hub, Keras.

Artificial Intelligent Art Project

Personal project on Artificial Intelligence with amazing Art.
☃ Explored the neural painter, a generative model for brushstrokes learned by a real non-differentiable and non-deterministic painting program.
Retrained the model in two parts: VAE to learn a latent space and a fully-connected network to map the action space to the latent space.
☕ Project Code.
Tools: Google Lucid, GAN, Python

System Modeling and Simulation

Independent Project.
2019.5 -- 2019.6)

☃Mastered classical data modeling approaches containing Lagrange Interpolation, Linear Interpolation, and Least Squares Fitting. Utilized MATLAB & Simulink with system simulation methods like Euler method, the improved trapezoidal method, fourth-order RK method to conduct simulation for several systems like Reusable launch vehicle system, interception system and pendulum system. ☕ Project Code.
Tools: MATLAB & Simulink

Image Classification on CIFAR-10

Personal project in computer vision.
☃Implemented image classification in computer vision with Tensorflow & Keras. Become familiar with data pre-processing, building great neural networks and classification on CIFAR-10 dataset.
Then utilized Tensorboard to inspect training processes and the machine learning model.
☕ Project Code.
Tools: Tensorflow, Python, Keras, Tensorboard.

More projects including self-driving car, reinforcement learning and virtual reality at ☕ My Github.


My Blogs

Technology // Minds // Travel

♛ Life is just like an amazing gem for us to make it much more brilliant.
☂ As I believe, if your beautiful heart holds hope, passions and love, you could harvest much more than what you want.
Furthermore, I would like to improve myself at any time and always encourage myself to try novel things in my life.
☘ As a 2nd-year master student who majors in Automation and Computer Science, undoubtedly I have fallen in love with advanced technology on Artificial Intelligence, Computer Vision, Machine Learning and Robotics.
♥ Not only mystery research programs, I also love Music (Blackpink [Jennie] & YOASOBI), Art, Fashion and Movies. My dream is traveling all around the world and implementing advanced Artificial Intelligence technology to make robot machines cook, sing, dance, draw and even generate excellent movies hhh.

2020.12.16 Finally, I opened my personal WeChat official account "Sarah星空日记" !! Welcome to scan the QR code and follow me. Feel glad to share you with my interesting life and ideas!