联系方式
- Email：yueyuebird.swjtu@gmail.com
个人信息
- 兴趣爱好：羽毛球/乒乓球/网球/骑行
- 本科/地理信息科学/西南交通大学/省级优秀毕业生
- 研究生/测绘科学与技术/西南交通大学/学硕在读
- 预计毕业时间：2025年6月
- 技术博客：https://yueyuebird-su.github.io/
- Github：https://github.com/YueyueBird-su
学科竞赛经历
国家级奖项
中国大学生服务外包创新创业大赛 企业命题类 二等奖
赛题“影像主体分割”，与遥感影像的建筑物提取等语义分割模型类似。我在该类项目中位次第二，在认真学习了一个暑假的pyqt后，完成了模型复现、训练、测试以及pyqt展示。后期在队友部署模型的基础上使用C++代码和QT5进行桌面端开发，并实现了一个“融合测试”的trick。解决了模型质量不行时导致的主体边缘不清晰的问题。 最终实现了一个对选择影像（文件夹）主体分割以及开启摄像头之后的实时分割并进行展示的软件。
省级奖项
全国大学生数学建模竞赛 四川省一等奖 两次
五一数学建模竞赛 二等奖/三等奖
在这些数学建模的赛事中，我主要负责将数学建模的同学所提出的模型用编程的思想解算，从而得到计算结果。例如1、多个力的相互作用；2、温度的传导，基本都是对常微分方程的解法。
科研经历
本科毕设 基于可变形卷积的多视影像密集匹配方法
完成了基于机载影像的多视角影像笔记匹配方法，采用一种可变形特征提取架构解决视角变化导致的特征形变的问题。
SRTP 基于拉曼光谱表征和1D卷积神经网络分类的水体污染物溯源
深度学习的启蒙项目。通过已有数据集对工厂水体排放的拉曼光谱进行分类，达到溯源的目的。其中主要工作是使用pyqt编写窗口。 最终优秀结题，并申请了一份软件著作权。
开源项目和作品
开源项目
正在努力中……
技术文章
技能清单
以下均为我熟练使用的技能
- C++面向对象编程：STL/Eigen/GDAL/Ceres/QT5
- Python：Pytorch/Numpy/PyQt/Pytorchlightning
-
- -
Contact Information.
- Email: yueyuebird.swjtu@gmail.com
Personal Information.
- Interests: Badminton/Table Tennis/Tennis/Cycling
- Bachelor’s Degree/Geographic Information Science/Southwest Jiaotong University/Outstanding Graduate of Province
- Master’s Degree/Geomatics Science and Technology/Southwest Jiaotong University/Current Student
- Expected Graduation Date: June 2025
- Technical Blog: https://yueyuebird-su.github.io/
- Github: https://github.com/YueyueBird-su
Competition Experience.
National Award.
China College Students Service Outsourcing Innovation and Entrepreneurship Competition, Enterprise Proposition, Second Prize.
The competition topic was “Image Subject Segmentation,” which is similar to semantic segmentation models such as building extraction from remote sensing images. I ranked second in this type of project and after carefully studying pyqt for a summer, I completed model reproduction, training, testing, and pyqt display. Later on, based on my teammate’s deployment of the model, I used C++ code and QT5 for desktop development and implemented a “fusion test” trick to solve the problem of unclear subject edges caused by poor model quality. Finally, I created software that can segment the main body of selected images (folders) and perform real-time segmentation and display after opening the camera.
Provincial Award National
College Students Mathematical Modeling Competition Sichuan Province, First Prize Twice
In these mathematical modeling competitions, I was mainly responsible for using programming ideas to solve the models proposed by my classmates and obtain calculation results. For example, 1. The interaction of multiple forces; 2. The conduction of temperature, which is basically the solution of ordinary differential equations
Research Experience
Undergraduate Thesis: Multi-view Image Dense Matching Method Based on Deformable Convolution
Completed a multi-view image dense matching method based on airborne images, using a deformable feature extraction architecture to solve the problem of feature deformation caused by changes in perspective.
SRTP: Water Pollution Source Tracing Based on Raman Spectroscopy Characterization and 1D Convolutional Neural Network Classification
A deep learning project. Classify Raman spectra of factory water discharge using an existing dataset to achieve source tracing. The main work was to write windows using pyqt. Finally, it was an excellent conclusion and applied for a software copyright.
Open Source Projects and Works
Open Source Projects
In progress…
Technical Articles
- Common Functions in Digital Image Processing
Skills
The following are the skills I am proficient in:
- C++ Object-Oriented Programming: STL/Eigen/GDAL/Ceres/QT5
- Python: Pytorch/Numpy/PyQt/Pytorchlightning
