Biography

Education

• Undergraduate (2016.08--2020.07): EE, Ocean University of China, with GPA = 3.28/4.0
• Master of Philosophy (2020.08--2023.06): CS, Ocean University of China, with GPA = 3.39/4.0
• 2016--2017: Second Class Scholarship for Study Excellent (Ranked 4th with 60)
• 2017--2018: Third Class Scholarship for Study Excellent (Ranked 7th with 60)
• 2018--2019: Scholarship for Science and Technology Innovation
• 2019--2020: Excellent Bachelor's Thesis
• 2022--2023: Graduate Academic Innovation Scholarship

Projects in Collage

• 2018.09--2019.06: Online Target Recognition System of Underwater Robot for Aquaculture

As the main completion of the project, the project was selected as the first innovation guidance special project of Ocean University of China and concluded with an excellent rating. The project deployed yolov3 (a target detection model) and deepsort (a target tracking algorithms) on the underwater robot to track and count underwater sea cucumber targets in real time.

The project initiated the ID reassignment algorithm, stored the target id and spatial information through the hash table, and determined again whether the target in a certain location was in the hash table of the previous frames in the tracking process of the following frames, which strengthened the ability to judge the same target, and improved the overall accuracy of underwater counting to 90%.

• 2022.03--2023.03: Real-time Wave Height Estimation System Based on Vision

A Natural Resources Bureau project which I in charge of as the student. Based on the shore-based fixed viewing angle camera, to estimate the real-time sea wave height in the visual area via integration of deep learning and triangle fitting method. Compared with the current mainstream wave measurement methods of radar (echo backthrust) and buoy (acceleration as integral) in the industry, this scheme is simple, direct and high accuracy. What is more outstanding is that this scheme can realize real-time wave height measurement and reflect to the software, while the current mainstream method in the industry can only take 15 minutes or even half an hour as the timing base.

Based on the above advantages, the scheme can also calculate the key parameters of wavelength, period, 1/3 wave height, 1/10 wave height, maximum and average wave height in a period of time and return to the background in real time.

Work

• 2021.09--2022.04: Algorithm Engineer, Qingdao Easy Touch Technology Co., LTD (Internship)

Responsible for the development and end-side deployment of unmanned vendor algorithms, converting the yolov5 target detection algorithm to a tensorRT inference model and deploying it to the JetsonNano. A pure IOU tracking algorithm is used to track whether an individual item is sold or put back. The whole process of detection, tracking and judgement runs in real-time, and the results can be returned to the service side within two seconds after the door closed. 93% comprehensive accuracy was achieved in indoor tests for 10 categories of goods.

• 2023.07--present: Device Software Engineer, NIO Autonomous Driving

Responsible for the Data Loopback (DLB) of the 2nd/3rd generation technology platform of NIO Autonomous Driving (NAD), managing the data (messages, files, etc. generated by sensors, algorithms, services, etc.) and data flow of the autonomous driving system in order to timely complete the interaction between vehicle and power station, the interaction between device and cloud, record the status of the vehicle and the autonomous driving system, and provide the necessary data support for the implementation and iteration of the autonomous driving algorithms.

Other Projects

• 2024.05--present: AI+CAD Intelligent Drawing Review System, Shanghai AI Laboratory

This CAD review and matching system is based on deep learning technology. It analyzes two or more images or CAD files stored in PDF format to determine whether they describe the same or similar design entities. The system is applied in design review, engineering re-examination, and the verification and validation of drawings in the manufacturing sector. It incorporates techniques such as few-shot learning, self-supervised learning, two-stage matching, incremental object detection using open datasets, lightweight neural network design, and network performance optimization for real-world deployment. The project is currently ongoing, with initial implementation based on visual analysis already achieved. Planned advancements include the development of a large-scale CAD visual model (for CAD understanding) and a structural understanding model combining graph neural networks. Ongoing research papers include submissions to Nature sub-journals and PAMI. This project is a collaboration between the Shanghai AI Laboratory and a startup company.

• 2024.07--present: OpenSTX: Open Smart Trading eXpert

This project is an open-source financial data quantitative analysis initiative using Interactive Brokers as the data source. It focuses on efficiently requesting, reorganizing, persistently storing, and feature engineering financial data (L1 and L2 level data for stocks and options). Advanced deep learning models are employed for quantitative analysis, model computation, visualization, automated metric generation, and incremental online learning to gain a deeper understanding of financial markets. The project aims to achieve real-time response to market data through a deep learning inference framework targeted at specific platforms (currently Apple Arm Silicon and MacOS Based). The goal of this project is to assist professionals in computer and statistical sciences in understanding financial markets through data, thereby lowering the barriers to entry for individuals with computer science and mathematics backgrounds. This is a personal project in its early stages, and the project repository is available at https://github.com/AilvenLiu/OpenSTX.

Competition

• 2019.05: Excellence Award in Future Cup University AI Challenge (4th place in North China)
• 2019.08: Second Prize of the Ninth Huawei Cup Intelligent Design Competition for College Students
• 2019.08: Second Prize of the Eighth National Marine Vehicle Design and Construction Competition
• 2020.12: Third Prize of the First Ocean Target Intelligent Sensing Challenge
• 2020.12: Grand Prize of National "IoT+" University Students' Science and Technology Innovation Competition
• 2021.08: Grand Prize of the 10th National Marine Vehicle Design and Construction Competition
• 2022.04: Third Prize of Huawei Software Elite Challenge in Jiangsu-Shandong Region

Research and Patents

• 2019: Incremental Layers Resection: A Novel Method to Compress Neural Networks, IEEE Access
• 2020: Compressing Deep Networks by Neuron Agglomerative Clustering, MDPI Sensors
• 2020: In Situ Holothurian Noncontact Counting System: A General Framework for Holothurian Counting, IEEE Access
• 2020: Merging Similar Neurons for Deep Networks Compression, Springer Cognitive Computation
• 2021: Deep Architecture Compression with Automatic Clustering of Similar Neurons, PRCV2021
• 2022: A Compact Object Detection Architecture with Transformer Enhancing, CMVIT2022
• 2022: SteelyGAN: Semantic Unsupervised Symbolic Music Genre Transfer, PRCV2022
• 2023: Quantum Prototype Clustering Algorithm, Computer Science (Chinese Journal)
• 2022: Lightweight merchandise identification tracking system, method, storage medium, device, terminal. CN115311450A
• 2022: A vending cabinet merchandise dynamic identification system, method, medium, device, and terminal. CN115294325A
• 2022: An Artificial Intelligence based wave video monitoring system. CN115880633A

Other experience

• 2018.09--2019.09: First Academic Minister of Intelligent Data Analytics Club, Ocean University of China
• 2016.08--2019.10: National Defence Student of People's Navy (North Sea Fleet), Ocean University of China
• 2018.12--2020.01: Part-time Barista of Starbucks
• 2019.08--2019.08: Exchange student at Pohang University of Science and Technology

Interest

Running, Tennis, Piano

Technology stack

Linux/C++, TensorRT/C++@Jetson, Pytorch

Comminucation

• Email: ailven.x.liu@gmail.com; ailven.liu@nio.com
• WeChat: OUC_LiuX

Talks

• Research Gate
• Google Scholar