svhn example

example_svhn_digit.py

+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+# File: example_svhn_digit.py
+# Author: Yuxin Wu <ppwwyyxx@gmail.com>
+
+import tensorflow as tf
+import argparse
+import numpy as np
+import os
+
+from tensorpack.train import TrainConfig, QueueInputTrainer
+from tensorpack.models import *
+from tensorpack.callbacks import *
+from tensorpack.utils import *
+from tensorpack.utils.symbolic_functions import *
+from tensorpack.utils.summary import *
+from tensorpack.dataflow import *
+from tensorpack.dataflow import imgaug
+
+
+class Model(ModelDesc):
+    def _get_input_vars(self):
+        return [InputVar(tf.float32, [None, 40, 40, 3], 'input'),
+                InputVar(tf.int32, [None], 'label') ]
+
+    def _get_cost(self, input_vars, is_training):
+        image, label = input_vars
+        keep_prob = tf.constant(0.5 if is_training else 1.0)
+
+        image = image / 255.0
+
+        nl = lambda x, name: tf.abs(tf.tanh(x), name=name)
+        l = Conv2D('conv1', image, 24, 5, padding='VALID', nl=nl)
+        l = MaxPooling('pool1', l, 2, padding='SAME')
+        l = Conv2D('conv2', l, 32, 3, nl=nl, padding='VALID')
+        l = Conv2D('conv3', l, 32, 3, nl=nl, padding='VALID')
+        l = MaxPooling('pool2', l, 2, padding='SAME')
+        l = Conv2D('conv4', l, 64, 3, nl=nl, padding='VALID')
+
+        l = tf.nn.dropout(l, keep_prob)
+        l = FullyConnected('fc0', l, 512,
+                           b_init=tf.constant_initializer(0.1), nl=nl)
+        # fc will have activation summary by default. disable for the output layer
+        logits = FullyConnected('linear', l, out_dim=10, summary_activation=False,
+                                nl=tf.identity)
+        prob = tf.nn.softmax(logits, name='output')
+
+        y = one_hot(label, 10)
+        cost = tf.nn.softmax_cross_entropy_with_logits(logits, y)
+        cost = tf.reduce_mean(cost, name='cross_entropy_loss')
+        tf.add_to_collection(MOVING_SUMMARY_VARS_KEY, cost)
+
+        # compute the number of failed samples, for ValidationError to use at test time
+        wrong = prediction_incorrect(logits, label)
+        nr_wrong = tf.reduce_sum(wrong, name='wrong')
+        # monitor training error
+        tf.add_to_collection(
+            MOVING_SUMMARY_VARS_KEY, tf.reduce_mean(wrong, name='train_error'))
+
+        # weight decay on all W of fc layers
+        wd_cost = tf.mul(0.00001,
+                         regularize_cost('fc.*/W', tf.nn.l2_loss),
+                         name='regularize_loss')
+        tf.add_to_collection(MOVING_SUMMARY_VARS_KEY, wd_cost)
+
+        add_param_summary([('.*/W', ['histogram', 'sparsity'])])   # monitor W
+        return tf.add_n([cost, wd_cost], name='cost')
+
+def get_config():
+    #anchors = np.mgrid[0:4,0:4][:,1:,1:].transpose(1,2,0).reshape((-1,2)) / 4.0
+    # prepare dataset
+    d1 = dataset.SVHNDigit('train')
+    d2 = dataset.SVHNDigit('extra')
+    train = RandomMixData([d1, d2])
+    test = dataset.SVHNDigit('test')
+
+    augmentors = [
+        imgaug.Resize((40, 40)),
+        imgaug.BrightnessAdd(63),
+        imgaug.Contrast((0.2,1.8)),
+    ]
+    train = AugmentImageComponent(train, augmentors)
+    train = BatchData(train, 128)
+    nr_proc = 2
+    train = PrefetchData(train, 3, nr_proc)
+    step_per_epoch = train.size() / nr_proc
+
+    augmentors = [
+        imgaug.Resize((40, 40)),
+    ]
+    test = AugmentImageComponent(test, augmentors)
+    test = BatchData(test, 128, remainder=True)
+
+    sess_config = get_default_sess_config()
+    sess_config.gpu_options.per_process_gpu_memory_fraction = 0.5
+
+    lr = tf.train.exponential_decay(
+        learning_rate=1e-4,
+        global_step=get_global_step_var(),
+        decay_steps=train.size() * 50,
+        decay_rate=0.7, staircase=True, name='learning_rate')
+    tf.scalar_summary('learning_rate', lr)
+
+    return TrainConfig(
+        dataset=train,
+        optimizer=tf.train.AdamOptimizer(lr),
+        callbacks=Callbacks([
+            StatPrinter(),
+            PeriodicSaver(),
+            ValidationError(test, prefix='test'),
+        ]),
+        session_config=sess_config,
+        model=Model(),
+        step_per_epoch=step_per_epoch,
+        max_epoch=100,
+    )
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--gpu', help='comma separated list of GPU(s) to use.') # nargs='*' in multi mode
+    parser.add_argument('--load', help='load model')
+    args = parser.parse_args()
+
+    basename = os.path.basename(__file__)
+    logger.set_logger_dir(
+        os.path.join('train_log', basename[:basename.rfind('.')]))
+
+    if args.gpu:
+        os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
+    else:
+        os.environ['CUDA_VISIBLE_DEVICES'] = '0'
+
+    with tf.Graph().as_default():
+        with tf.device('/cpu:0'):
+            config = get_config()
+        if args.load:
+            config.session_init = SaverRestore(args.load)
+        if args.gpu:
+            config.nr_tower = len(args.gpu.split(','))
+        QueueInputTrainer(config).train()

tensorpack/dataflow/dataset/svhn.py

+#!/usr/bin/env python2
+# -*- coding: UTF-8 -*-
+# File: svhn.py
+# Author: Yuxin Wu <ppwwyyxx@gmail.com>
+
+import os
+import random
+import numpy
+import scipy
+import scipy.io
+from six.moves import range
+
+from ...utils import logger
+from ..base import DataFlow
+
+__all__ = ['SVHNDigit']
+
+class SVHNDigit(DataFlow):
+    """
+    SVHN Cropped Digit Dataset
+    return img of 32x32x3, label of 0-9
+    """
+    def __init__(self, name, data_dir=None):
+        """
+        name: 'train', 'test', or 'extra'
+        """
+        if data_dir is None:
+            data_dir = os.path.join(
+                os.path.dirname(__file__),
+                'svhn_data')
+        assert name in ['train', 'test', 'extra'], name
+        filename = os.path.join(data_dir, name + '_32x32.mat')
+        assert os.path.isfile(filename), \
+                "File {} not found! Download it from \
+http://ufldl.stanford.edu/housenumbers/".format(filename)
+        logger.info("Loading {} ...".format(filename))
+        data = scipy.io.loadmat(filename)
+        self.X = data['X'].transpose(3,0,1,2)
+        self.Y = data['y'].reshape((-1))
+        self.Y[self.Y==10] = 0
+
+    def size(self):
+        return self.X.shape[0]
+
+    def get_data(self):
+        n = self.X.shape[0]
+        for k in range(n):
+            yield [self.X[k], self.Y[k]]
+