https://github.com/dizcza/ujipen
UJIPEN2 classification with Gated Recurrent Unit Neural Network.
https://github.com/dizcza/ujipen
dtw gru handwriting-recognition
Last synced: 8 months ago
JSON representation
UJIPEN2 classification with Gated Recurrent Unit Neural Network.
- Host: GitHub
- URL: https://github.com/dizcza/ujipen
- Owner: dizcza
- License: mit
- Created: 2019-02-21T11:19:50.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2024-05-03T20:11:06.000Z (about 2 years ago)
- Last Synced: 2025-09-27T05:55:11.549Z (9 months ago)
- Topics: dtw, gru, handwriting-recognition
- Language: Python
- Homepage:
- Size: 50.8 KB
- Stars: 1
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# UjipenChars2 handwritten letters classifier with Gated Recurrent Unit (GRU)
This repository supplements [stm32f429-chars](https://github.com/dizcza/stm32f429-chars) repository to train a recurrent neural network that will be used later on in a microcontroller.
A small list of manually picked examples from train data which confuse classifiers is put in [`dropped.txt`](ujipenchars2/dropped.txt). All test samples from [UjipenChars2](https://archive.ics.uci.edu/ml/datasets/UJI+Pen+Characters+(Version+2)) dataset are used during the model validation.
The main file is [`gru.py`](gru.py), where the training procedure of GRU is defined alongside with the test (validation) score.
Initially started with DTW as a baseline algorithm to find the closest pattern from the train data, given an input sample. DTW-related implementation is moved to [dtw](https://github.com/dizcza/ujipen/tree/dtw) branch.
To give you the rough approximation of performance of both classifiers,
| | GRU | DTW |
|---------------------|--------|--------|
| Validation accuracy | 98.3 % | 81.9 % |
But the main difference between those two is their inference time: GRU is much faster than DTW due to parallel computation.