import numpy as np

from os.path import expanduser, join

import os

import json

import theano

import pyret.filtertools as ft

import pyret.visualizations as pyviz

from preprocessing import datagen, loadexpt

from utils import rolling_window

from keras.models import model_from_json

import h5py

import matplotlib.pyplot as plt

from utils import mksavedir

save_dir = mksavedir(prefix='Contrast Steps')

# GET LN AND CNN RESPONSES TO CONTRAST STEPS

architecture_filename = 'architecture.json'

naturalscenes_data_dir = expanduser('~/Dropbox/deep-retina/saved/lenna.salamander/2015-11-07 16.52.44 convnet/')

naturalscenes_weight_filename = 'epoch038_iter02700_weights.h5' # .53 cc on held-out

ln_data_dir = expanduser('~/Dropbox/deep-retina/saved/lenna.nirum/2015-11-08 04.41.18 LN/')

ln_weight_filename = 'epoch010_iter00750_weights.h5' # .468 cc on held-out

# LOAD NATURAL SCENES MODEL

naturalscenes_architecture_data = open(naturalscenes_data_dir + architecture_filename, 'r')

naturalscenes_architecture_string = naturalscenes_architecture_data.read()

naturalscenes_model = model_from_json(naturalscenes_architecture_string)

naturalscenes_model.load_weights(naturalscenes_data_dir + naturalscenes_weight_filename)

# LOAD LN MODEL

ln_architecture_data = open(ln_data_dir + architecture_filename, 'r')

ln_architecture_string = ln_architecture_data.read()

ln_model = model_from_json(ln_architecture_string)

ln_model.load_weights(ln_data_dir + ln_weight_filename)

def get_full_field_flicker(period=1, low_contrast=0.1, high_contrast=1.0):

return full_field_movies

ntrials = 10000

period = 1

sample_rate = 100

batch_size = 26

nsamples = 3*period*sample_rate - 40

assert nsamples % batch_size == 0, 'nsamples must be divisible by batch_size'

cnn_responses = np.zeros((ntrials, nsamples))

ln_responses = np.zeros((ntrials, nsamples))

for n in range(ntrials):

if n % 50 == 0:

print 'Starting %d trial out of %d.' %(n,ntrials)

stimulus = get_full_field_flicker()

for batch in range(nsamples/batch_size):

cnn_responses[n, batch*batch_size:(batch+1)*batch_size] = \

naturalscenes_model.predict(stimulus[batch*batch_size:(batch+1)*batch_size])[:,0]

ln_responses[n, batch*batch_size:(batch+1)*batch_size] = \

ln_model.predict(stimulus[batch*batch_size:(batch+1)*batch_size])[:,0]

f = h5py.File(join(save_dir, 'responses_to_contrast_steps.h5'), 'w')

f.create_dataset('ln_responses', data=ln_responses)

f.create_dataset('cnn_responses', data=cnn_responses)

f.close()

fig = plt.gcf()

fig.set_size_inches((10,8))

# Plot multiple flicker sequences

for flicker in flicker_sequences:

plt.plot(np.linspace(0,2.6,260), 14+flicker[40:], 'k')

# Plot average CNN and LN responses

plt.plot(np.linspace(0.0,2.6,260), average_cnn_response, 'b', linewidth=3)

plt.plot(np.linspace(0.0,2.6,260), average_ln_response, 'r', linewidth=3)

plt.xlabel('Time (seconds)', fontsize=20)

#plt.title('RED: LN Responses, BLUE: CNN Responses', fontsize=20)

plt.savefig(join(save_dir, 'Contrast Adaptation - white noise LN model - top.png'), dpi=300)

fig = plt.gcf()

fig.set_size_inches((5,8))

# Plot multiple flicker sequences

for flicker in flicker_sequences:

plt.plot(np.linspace(0,2,200), 14+flicker[50+50:50+250], 'k')

# Plot average CNN and LN responses

plt.plot(np.linspace(0.0,2,200), average_cnn_response[10+50:10+250], 'b', linewidth=3)

plt.plot(np.linspace(0.0,2,200), average_ln_response[10+50:10+250], 'r', linewidth=3)

plt.xlabel('Time (seconds)', fontsize=20)

#plt.title('RED: LN Responses, BLUE: CNN Responses', fontsize=20)

plt.savefig(join(save_dir, 'Contrast Adaptation - white noise LN model - cropped - top.png'), dpi=300)