Projects

I've participated in DJI Robomaster Competition, National Training Program of Innovation and Entrepreneurship for Undergraduates, some course_based researches and other projects.

2022

> A new Method on Aircraft Attitude Adjustment based on Event-Camera

Final Codes

This is a national level project from National Training Program of National Training Program of Innovation and Entrepreneurship for Undergraduates, we use the newly promoted Event Camera as star sensor to get the attitude information with the help of stars in the space.

 

> Graph Visualization with pyecharts

xlsx2json & graph making

  Echarts is an open source library from Baidu for data visualization in javascript. It has awesome demo pages. I found this useful repositry and use it to visualize the Social Network Graph of the class of Computational Social Science course.

blog: pyechart实现可视化

 

> Tuning Pytorch Training with HUAWEI OpenEuler A-Tune engine

Training without A-Tune | training using A-Tune

  A-Tune is an OS tuning engine based on AI. I'm extremely surprised by this idea of using operating system to tuning services for the Users. I made an example of using A-Tune to tuning the batch-size, optimizer and learning-rate of the pytorch training (taking the CIFAR-10 DATASET as an example). The result is quite cool that it outperforms the result without tuning by 6.35% in 30 turns.

2021

> DJI RoboMaster | Engineer Robot Automatic Target Recognition and Pose Refining

  

real-time PCL Code Repositry

  DJI RoboMaster is truly an amazing competition for collegers. I believe the experience preparing for this competition with my teamates will absolutly be the most unforgetable ones during my life time. In this project, I use Computer Vision algorithms to regognize the featured items in the environment(the golden block in the picture) and use Robot Control algorithms and ICT designing to let the robot automatically refine its pose to grab the golden block.

 

> domestic waste Auto Classification Robot

  

   this is a smart waste can that is able to automatically classify different kind of domestic waste and push them into the specific basket. The core ability of classification is powered by DarkNet, which is trained on the manual collected dataset.