Embedded Files
An EECS C106A-19Fall project produced by
Group-3 members: Gan-lin Zhang, Teng Xu, Wei-jie Lyu, Zhen-zhong Tang, Zi-yuan Hu
at UC Berkeley.
Source code on Github
Our Goals
Our Goals
- Design a control system that allows a robot to draw a portrait / write Chinese characters.
- Design a path-finding algorithm that could generate a path to draw a portrait in one stroke.
- Implement a neural network that is capable of performing Chinese handwriting recognition tasks and generate the path for characters.
Extension
Extension
- This project would be useful for any arm-robots with at least 4 joints.
Spotlight
Spotlight
- Self designed coordinate transformation, avoiding unpredictable factors from Jacobian.
- Self designed path planner.
- Modified U-Net, i.e. LikeUnet, for human matting, with small model size, fast runtime, and high accuracy.
- Modified the handwritten digits recognition neural network Lenet-5 to make it suitable for Chinese handwriting recognition.
Page updated
Google Sites
Report abuse