Recurrent Neural Network with Connectionist Temporal Classifical implemented in Theano. Includes a Toy training example.
First generate some data using one of the scribes (a, b or c)
# Run with '-h' to see full functionality of gen_data.py
python3 gen_data.py -ha) Hindu numerals
python3 gen_data.py data.pkl -a hindub) ASCII characters.
python3 gen_data.py data.pkl -a asciic) scribe_rows scribes i-th digit along the i-th row as an i+2 long bar
python3 alphabets/scribe_rows.py data.pkl
# Run with no arguments to see full usageNow you have the data.pkl file.
Run the actual Recurrent Neural Net with Connectionist Temporal Classification cost function as:
python3 train.py data.pkl [configuration_num]
# Run with no arguments for full usage.# Using data from scribe_rows.py
Shown : 0 2 3 1 0 1 0 2 1 2
Seen : 0 2 3 1 0 1 0 2 1 2
Images (Shown & Seen) :
0¦ ██ ██ ██ ¦
1¦ ███ ███ ███ ¦
2¦ ████ ████ ████ ¦
3¦ █████ ¦
0¦ █ ██ █▓ ¦
1¦ ██ █ ███ ¦
2¦ █ ▒██ ██▓ ¦
3¦ █ ¦
4¦████ ▒████¦
# Using data from scribe.py hindu
Shown : 0 2 2 5
Seen : 0 2 2 5
Images (Shown & Seen) :
0¦ ¦
1¦ ██ ██ ¦
2¦ █ ██ ████ ¦
3¦ █ █ █ ¦
4¦ ██ █ █ ███ ¦
5¦ █ █████████ █ ¦
6¦ █ █ █ █ ¦
7¦ ██ ███ ¦
0¦░░░░░░░░░█░░░░░░░░░░░░░░░░░░¦
1¦░░░░░░░░░░░░░░░░░░░░░░░░░░░░¦
2¦░░░░░░░░░░░░░█░░░█░░░░░░░░░░¦
3¦░░░░░░░░░░░░░░░░░░░░░░░░░░░░¦
4¦░░░░░░░░░░░░░░░░░░░░░░░░░░░░¦
5¦░░░░░░░░░░░░░░░░░░░█▓░░░░░░░¦
6¦█████████░███░███░█░▒███████¦
- Graves, Alex. Supervised Sequence Labelling with Recurrent Neural Networks. Chapters 2, 3, 7 and 9.
- Available at Springer
- University Edition via. Springer Link.
- Free Preprint
- Theano implementation of CTC by Shawn Tan
- Numpy
- Theano
Can easily port to python2 by adding lines like these where necessary:
from __future__ import print_function
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent