Pytorch Handwritten Mathematical Expression Recognition
This program uses Attention and Coverage to realize HMER and this program is based on Pytorch.
Overview:
The Handwritten-Mathematical-Expression-Recognition (Pytorch) program utilizes Attention and Coverage to achieve Handwritten Mathematical Expression Recognition (HMER). It was developed by Hongyu Wang with guidance from Dr. Jianshu Zhang. The README.md file was updated on August 13. The program is designed for recognition tasks and provides impressive results in the CROHME 2016 dataset when tested on two TITAN XP GPUs.
Features:
- Utilizes Attention and Coverage for HMER
- Training and Testing capabilities
- Pretrained Densenet weights available for download
- Data compression into '.pkl' files
- Batch size of 6, max length of 48, and max image size of 100,000
- Achieved best results with WER loss at 17.160% and ExpRate at 38.595%
- Visualization of results and attention mechanism
Summary:
The Handwritten-Mathematical-Expression-Recognition program in Pytorch, developed by Hongyu Wang, offers a robust solution for recognizing handwritten mathematical expressions. By incorporating Attention and Coverage mechanisms, the program achieves remarkable results as demonstrated in the CROHME 2016 dataset. The program's features include data compression, visualization of results, and efficient training and testing processes. For those interested in exploring the realm of Handwritten Mathematical Expression Recognition, this program provides a solid foundation and impressive results to build upon.