This repository contains our experiments with the deep learning library Keras and a subset of the CASIA-HWDB1.1 data set. The code here provided implements a convolutional neural network with test set accuracy sometimes greater than 95%. The network configuration is a simplified version of the one described in Deep Convolutional Network for Handwritten Chinese Character Recognition by Yuhao Zhang; no attempt was made to improve the model since we just wanted to play with convolutional networks and deep learning libraries.
Alessandro and Francesco
We assume you have already installed and configured Keras with any of its backends. Install the other required dependencies with:
$ pip3 install -r requirements.txt
The code was last tested on Keras 2.1.5 using TensorFlow 1.8.0 as backend, h5py 2.7.1, numpy 1.14.2, Pillow 5.1.0, scikit-image 0.14.0 and scipy 1.0.0. Tests were run under a single GeForce GTX 970.
In the releases section we uploaded a (zipped) subset of the CASIA-HWDB1.1 data set, a trained model and some classifications, all generated following the steps below. If you use our subset, start from step 3.
- Download the CASIA-HWDB1.1 data set from the official locations (HWDB1.1trn_gnt.zip (1873 MB) and HWDB1.1tst_gnt.zip (471 MB)) and unzip it (it is required to decompress an archive in ALZ format):
$ unzip HWDB1.1trn_gnt.zip
Archive: HWDB1.1trn_gnt.zip
inflating: HWDB1.1trn_gnt.alz
$ unalz HWDB1.1trn_gnt.alz
unalz v0.65 (2009/04/01)
Copyright(C) 2004-2009 by [email protected] (http://www.kipple.pe.kr)
Extract HWDB1.1trn_gnt.alz to .
unalziiiing : 1219-c.gnt (21547829bytes) ...........
unalziiiing : 1220-c.gnt (26773966bytes) ...........
[...]
$ mkdir HWDB1.1trn_gnt
$ mv *.gnt HWDB1.1trn_gnt/
$ unzip HWDB1.1tst_gnt.zip
Archive: HWDB1.1tst_gnt.zip
inflating: 1289-c.gnt
inflating: 1290-c.gnt
[...]
$ mkdir HWDB1.1tst_gnt
$ mv *.gnt HWDB1.1tst_gnt/
- Convert the data set into the HDF5 binary data format:
$ python3 1-gnt_to_dataset.py HWDB1.1trn_gnt/ HWDB1.1tst_gnt/
Converting 'trn'...
Converting 'tst'...
- Extract from the HDF5 data set a subset of 200 classes of characters:
$ python3 2-dataset_to_subset.py HWDB1.1.hdf5
Subsetting 'trn'...
Subsetting 'tst'...
- Train the network on the subset:
$ python3 3-train_subset.py HWDB1.1subset.hdf5
Using TensorFlow backend.
. . .
2018-06-10 10:18:51.390792: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 3608 MB memory) -> physical GPU (device: 0, name: GeForce GTX 970, p
ci bus id: 0000:02:00.0, compute capability: 5.2)
Train on 40000 samples, validate on 8020 samples
Epoch 1/15
40000/40000 [==============================] - 41s 1ms/step - loss: 4.5114 - acc: 0.0975 - val_loss: 2.2487 - val_a
cc: 0.4542
Epoch 2/15
40000/40000 [==============================] - 30s 761us/step - loss: 1.7218 - acc: 0.5745 - val_loss: 0.7780 - val
_acc: 0.8026
Epoch 3/15
40000/40000 [==============================] - 30s 762us/step - loss: 0.8618 - acc: 0.7704 - val_loss: 0.4930 - val
_acc: 0.8675
. . .
Epoch 14/15
40000/40000 [==============================] - 31s 763us/step - loss: 0.1265 - acc: 0.9607 - val_loss: 0.3307 - val_acc: 0.9181
Epoch 15/15
40000/40000 [==============================] - 30s 762us/step - loss: 0.1179 - acc: 0.9640 - val_loss: 0.3282 - val_acc: 0.9198
Test score: 0.32826207710256655
Test accuracy: 0.9214865656650205
- (Optional) Generate a report of some classifications:
$ python3 4-draw_results.py HWDB1.1subset.hdf5 model-1528618732.json weights-1528618732-0.921487.hdf5
Using TensorFlow backend.
. . .
2018-06-10 10:33:17.881820: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 3608 MB memory) -> physical GPU (device: 0, name: GeForce GTX 970, pci bus id: 0000:02:00.0, compute capability: 5.2)
Evaluating the network on the test set...
Test score: 0.32826207710256655
Test accuracy: 0.9214865656650205
Extracting some results...