final_attempt_pretrain.py

# This program pre-trains a saliency model on ground truths extracted from ImageNet database
# Input: RGB images and corresponding fixation maps on ImageNet
# Output: A trained CNN model(which will be used during 2nd phase training) for visual saliency on ImageNet 



import cPickle as pickle
from datetime import datetime
import os
import sys
import scipy
from scipy import misc

from matplotlib import pyplot
import numpy as np
from lasagne import layers
from nolearn.lasagne import BatchIterator
from nolearn.lasagne import NeuralNet
from pandas import DataFrame
from pandas.io.parsers import read_csv
from sklearn.utils import shuffle
import theano
from lasagne import layers
from lasagne.updates import nesterov_momentum
from nolearn.lasagne import NeuralNet
from nolearn.lasagne import BatchIterator	
from sklearn.externals import joblib
from sklearn.base import clone
import scipy.io as sio


image_size=128
map_size=32

def float32(k):
    return np.cast['float32'](k)

class AdjustVariable(object):
    def __init__(self, name, start=0.03, stop=0.001):
        self.name = name
        self.start, self.stop = start, stop
        self.ls = None

    def __call__(self, nn, train_history):
        if self.ls is None:
            self.ls = np.linspace(self.start, self.stop, nn.max_epochs)

        epoch = train_history[-1]['epoch']
        new_value = float32(self.ls[epoch - 1])
        getattr(nn, self.name).set_value(new_value)


pretrain_net = NeuralNet(
    layers=[
        ('input', layers.InputLayer),
        ('conv1', layers.Conv2DLayer),
        ('pool1', layers.MaxPool2DLayer),
        ('conv2', layers.Conv2DLayer),
        ('pool2', layers.MaxPool2DLayer),
        ('conv3', layers.Conv2DLayer),
        ('pool3', layers.MaxPool2DLayer),
        ('hidden4', layers.DenseLayer),
        ('maxout6',layers.FeaturePoolLayer),
        ('output', layers.DenseLayer),
        ],
    input_shape=(None, 3, image_size, image_size),
    conv1_num_filters=map_size, conv1_filter_size=(5, 5), pool1_pool_size=(2, 2),
    conv2_num_filters=64, conv2_filter_size=(3, 3), pool2_pool_size=(2, 2),
    conv3_num_filters=64, conv3_filter_size=(3, 3), pool3_pool_size=(2, 2),
    hidden4_num_units=map_size*map_size*2,
    maxout6_pool_size=2,output_num_units=map_size*map_size,output_nonlinearity=None,

    update_learning_rate=theano.shared(float32(0.05)),
    update_momentum=theano.shared(float32(0.9)),

    regression=True,

    on_epoch_finished=[
        AdjustVariable('update_learning_rate', start=0.05, stop=0.0001),
        AdjustVariable('update_momentum', start=0.9, stop=0.999),
        ],
    batch_iterator_train=BatchIterator(batch_size=128),
    max_epochs=320,
    verbose=1,
    )

no_of_examples=np.load('no_of_example.npy')
Xtr = np.memmap('X4.npy', dtype='float32', mode='r', shape=(no_of_examples,3,image_size,image_size))
ytr = np.memmap('y4.npy', dtype='float32', mode='r', shape=(no_of_examples,map_size*map_size))
print 'During training shape of X and y',Xtr.shape, ytr.shape
pretrain_net.fit(Xtr, ytr)
joblib.dump(pretrain_net, 'pretrain_net.pkl')
###############################################################################################3