def load_network(loadfile, multiview=None, sort_layers=False):
from convdata import DataProvider
from python_util.gpumodel import IGPUModel
load_dic = IGPUModel.load_checkpoint(loadfile)
layers = load_dic['model_state']['layers']
op = load_dic['op']
if sort_layers:
depths = get_depths(layers)
layers = collections.OrderedDict(
sorted(layers.items(), key=lambda item: depths[item[0]]))
options = {}
for o in load_dic['op'].get_options_list():
options[o.name] = o.value
dp_params = {}
for v in ('color_noise', 'multiview_test', 'inner_size', 'scalar_mean',
'minibatch_size'):
dp_params[v] = options[v]
lib_name = "cudaconvnet._ConvNet"
print("Importing %s C++ module" % lib_name)
libmodel = __import__(lib_name, fromlist=['_ConvNet'])
dp_params['libmodel'] = libmodel
if multiview is not None:
dp_params['multiview_test'] = multiview
dp = DataProvider.get_instance(options['data_path'],
batch_range=options['test_batch_range'],
type=options['dp_type'],
dp_params=dp_params,
test=True)
epoch, batchnum, data = dp.get_next_batch()
images, labels = data[:2]
images = images.T
images.shape = (images.shape[0], dp.num_colors, dp.inner_size,
dp.inner_size)
labels.shape = (-1, )
labels = labels.astype('int')
assert images.shape[0] == labels.shape[0]
if 1:
rng = np.random.RandomState(8)
i = rng.permutation(images.shape[0])
images = images[i]
labels = labels[i]
data = [images, labels] + list(data[2:])
# data['data_mean'] = dp.data_mean
# data['label_names'] = dp.batch_meta['label_names']
return layers, data, dp
def plain_pickle(loadfile, savefile):
load_dic = IGPUModel.load_checkpoint(loadfile)
options = {}
for o in load_dic['op'].get_options_list():
options[o.name] = o.value
load_dic['op'] = options
with open(savefile, 'wb') as f:
pickle.dump(load_dic, f, protocol=-1)
print("Wrote %r" % savefile)
""" View the options used to create a checkpoint
python view_options.py --load-file <checkpoint>
"""
from convnet import ConvNet
from python_util.gpumodel import IGPUModel
op = ConvNet.get_options_parser()
op, load_dic = IGPUModel.parse_options(op)
model = ConvNet(op, load_dic)
model.op.print_values()
print "========================="
model.print_model_state()
def loadbcp(name, shape, params=None):
load_dic = IGPUModel.load_checkpoint(params[0])
rows, cols = shape
return load_dic['model_state']['layers'][params[1]]['biases'].reshape(rows,cols)
def loadwcp(name, idx, shape, params=None):
load_dic = IGPUModel.load_checkpoint(params[0])
rows, cols = shape
return load_dic['model_state']['layers'][params[1]]['weights'][idx].reshape(rows,cols)
def loadbcp(name, shape, params=None):
load_dic = IGPUModel.load_checkpoint(params[0])
rows, cols = shape
return load_dic['model_state']['layers'][params[1]]['biases'].reshape(
rows, cols)
def loadwcp(name, idx, shape, params=None):
load_dic = IGPUModel.load_checkpoint(params[0])
rows, cols = shape
return load_dic['model_state']['layers'][
params[1]]['weights'][idx].reshape(rows, cols)
if not donames:
load_convnets = sys.argv[1:] # just the paths
else:
load_convnets = sys.argv[1::2]
legend_names = sys.argv[2::2] # the paths and a shorter legend name
print 'Loading convnets ' + ', '.join(load_convnets)
for cost_idx in range(ncosts):
pl.figure(cost_idx + basefig)
if navg_groups > 1:
allte = None
alltr = None
for cv, i in zip(load_convnets, range(len(load_convnets))):
cp = IGPUModel.load_checkpoint(cv)
numbatches, train_errors, test_errors = get_errors(
cp,
show_cost,
cost_idx,
norm=cost_idx == 0,
smooth=smooth,
interp_test=dointerp)
#plot_errors(numbatches, train_errors, test_errors)
tr = np.array(train_errors, dtype=np.double)
te = np.array(test_errors, dtype=np.double)
if navg_groups > 1:
if allte is None:
allte = np.zeros(list(te.shape) + [len(load_convnets)],
dtype=np.double)
alltr = np.zeros(list(tr.shape) + [len(load_convnets)],
def plot_predictions(self):
epoch, batch, data = self.get_next_batch(train=False) # get a test batch
num_classes = self.test_data_provider.get_num_classes()
NUM_ROWS = 2
NUM_COLS = 4
NUM_IMGS = NUM_ROWS * NUM_COLS if not self.save_preds else data[0].shape[1]
NUM_TOP_CLASSES = min(num_classes, 5) # show this many top labels
NUM_OUTPUTS = self.model_state["layers"][self.softmax_name]["outputs"]
PRED_IDX = 1
label_names = [lab.split(",")[0] for lab in self.test_data_provider.batch_meta["label_names"]]
if self.only_errors:
preds = n.zeros((data[0].shape[1], NUM_OUTPUTS), dtype=n.single)
else:
preds = n.zeros((NUM_IMGS, NUM_OUTPUTS), dtype=n.single)
# rand_idx = nr.permutation(n.r_[n.arange(1), n.where(data[1] == 552)[1], n.where(data[1] == 795)[1], n.where(data[1] == 449)[1], n.where(data[1] == 274)[1]])[:NUM_IMGS]
rand_idx = nr.randint(0, data[0].shape[1], NUM_IMGS)
if NUM_IMGS < data[0].shape[1]:
data = [n.require(d[:, rand_idx], requirements="C") for d in data]
# data += [preds]
# Run the model
print [d.shape for d in data], preds.shape
self.libmodel.startFeatureWriter(data, [preds], [self.softmax_name])
IGPUModel.finish_batch(self)
print preds
data[0] = self.test_data_provider.get_plottable_data(data[0])
if self.save_preds:
if not gfile.Exists(self.save_preds):
gfile.MakeDirs(self.save_preds)
preds_thresh = preds > 0.5 # Binarize predictions
data[0] = data[0] * 255.0
data[0][data[0] < 0] = 0
data[0][data[0] > 255] = 255
data[0] = n.require(data[0], dtype=n.uint8)
dir_name = "%s_predictions_batch_%d" % (os.path.basename(self.save_file), batch)
tar_name = os.path.join(self.save_preds, "%s.tar" % dir_name)
tfo = gfile.GFile(tar_name, "w")
tf = TarFile(fileobj=tfo, mode="w")
for img_idx in xrange(NUM_IMGS):
img = data[0][img_idx, :, :, :]
imsave = Image.fromarray(img)
prefix = (
"CORRECT"
if data[1][0, img_idx] == preds_thresh[img_idx, PRED_IDX]
else "FALSE_POS"
if preds_thresh[img_idx, PRED_IDX] == 1
else "FALSE_NEG"
)
file_name = "%s_%.2f_%d_%05d_%d.png" % (
prefix,
preds[img_idx, PRED_IDX],
batch,
img_idx,
data[1][0, img_idx],
)
# gf = gfile.GFile(file_name, "w")
file_string = StringIO()
imsave.save(file_string, "PNG")
tarinf = TarInfo(os.path.join(dir_name, file_name))
tarinf.size = file_string.tell()
file_string.seek(0)
tf.addfile(tarinf, file_string)
tf.close()
tfo.close()
# gf.close()
print "Wrote %d prediction PNGs to %s" % (preds.shape[0], tar_name)
else:
fig = pl.figure(3, figsize=(12, 9))
fig.text(0.4, 0.95, "%s test samples" % ("Mistaken" if self.only_errors else "Random"))
if self.only_errors:
# what the net got wrong
if NUM_OUTPUTS > 1:
err_idx = [i for i, p in enumerate(preds.argmax(axis=1)) if p not in n.where(data[2][:, i] > 0)[0]]
else:
err_idx = n.where(data[1][0, :] != preds[:, 0].T)[0]
print err_idx
err_idx = r.sample(err_idx, min(len(err_idx), NUM_IMGS))
data[0], data[1], preds = data[0][:, err_idx], data[1][:, err_idx], preds[err_idx, :]
import matplotlib.gridspec as gridspec
import matplotlib.colors as colors
cconv = colors.ColorConverter()
gs = gridspec.GridSpec(NUM_ROWS * 2, NUM_COLS, width_ratios=[1] * NUM_COLS, height_ratios=[2, 1] * NUM_ROWS)
# print data[1]
for row in xrange(NUM_ROWS):
for col in xrange(NUM_COLS):
img_idx = row * NUM_COLS + col
if data[0].shape[0] <= img_idx:
break
pl.subplot(gs[(row * 2) * NUM_COLS + col])
# pl.subplot(NUM_ROWS*2, NUM_COLS, row * 2 * NUM_COLS + col + 1)
pl.xticks([])
pl.yticks([])
img = data[0][img_idx, :, :, :]
img = img.squeeze()
if len(img.shape) > 2: # more than 2 dimensions
if img.shape[2] is 2: # if two channels
# copy 2nd to 3rd channel for visualization
a1 = img
a2 = img[:, :, 1]
a2 = a2[:, :, n.newaxis]
img = n.concatenate((a1, a2), axis=2)
pl.imshow(img, interpolation="lanczos")
else:
pl.imshow(img, interpolation="lanczos", cmap=pl.gray())
show_title = data[1].shape[0] == 1
true_label = [int(data[1][0, img_idx])] if show_title else n.where(data[1][:, img_idx] == 1)[0]
# print true_label
# print preds[img_idx,:].shape
# print preds[img_idx,:].max()
true_label_names = [label_names[i] for i in true_label]
img_labels = sorted(zip(preds[img_idx, :], label_names), key=lambda x: x[0])[-NUM_TOP_CLASSES:]
# print img_labels
axes = pl.subplot(gs[(row * 2 + 1) * NUM_COLS + col])
height = 0.5
ylocs = n.array(range(NUM_TOP_CLASSES)) * height
pl.barh(
ylocs,
[l[0] for l in img_labels],
height=height,
color=["#ffaaaa" if l[1] in true_label_names else "#aaaaff" for l in img_labels],
)
# pl.title(", ".join(true_labels))
if show_title:
pl.title(", ".join(true_label_names), fontsize=15, fontweight="bold")
else:
print true_label_names
pl.yticks(
ylocs + height / 2,
[l[1] for l in img_labels],
x=1,
backgroundcolor=cconv.to_rgba("0.65", alpha=0.5),
weight="bold",
)
for line in enumerate(axes.get_yticklines()):
line[1].set_visible(False)
# pl.xticks([width], [''])
# pl.yticks([])
pl.xticks([])
pl.ylim(0, ylocs[-1] + height)
pl.xlim(0, 1)
def plot_predictions(self):
epoch, batch, data = self.get_next_batch(
train=False) # get a test batch
num_classes = self.test_data_provider.get_num_classes()
NUM_ROWS = 2
NUM_COLS = 4
NUM_IMGS = NUM_ROWS * NUM_COLS if not self.save_preds else data[
0].shape[1]
NUM_TOP_CLASSES = min(num_classes, 5) # show this many top labels
NUM_OUTPUTS = self.model_state['layers'][self.softmax_name]['outputs']
PRED_IDX = 1
label_names = [
lab.split(',')[0]
for lab in self.test_data_provider.batch_meta['label_names']
]
if self.only_errors:
preds = n.zeros((data[0].shape[1], NUM_OUTPUTS), dtype=n.single)
else:
preds = n.zeros((NUM_IMGS, NUM_OUTPUTS), dtype=n.single)
#rand_idx = nr.permutation(n.r_[n.arange(1), n.where(data[1] == 552)[1], n.where(data[1] == 795)[1], n.where(data[1] == 449)[1], n.where(data[1] == 274)[1]])[:NUM_IMGS]
rand_idx = nr.randint(0, data[0].shape[1], NUM_IMGS)
if NUM_IMGS < data[0].shape[1]:
data = [
n.require(d[:, rand_idx], requirements='C') for d in data
]
# data += [preds]
# Run the model
print[d.shape for d in data], preds.shape
self.libmodel.startFeatureWriter(data, [preds], [self.softmax_name])
IGPUModel.finish_batch(self)
print preds
data[0] = self.test_data_provider.get_plottable_data(data[0])
if self.save_preds:
if not gfile.Exists(self.save_preds):
gfile.MakeDirs(self.save_preds)
preds_thresh = preds > 0.5 # Binarize predictions
data[0] = data[0] * 255.0
data[0][data[0] < 0] = 0
data[0][data[0] > 255] = 255
data[0] = n.require(data[0], dtype=n.uint8)
dir_name = '%s_predictions_batch_%d' % (os.path.basename(
self.save_file), batch)
tar_name = os.path.join(self.save_preds, '%s.tar' % dir_name)
tfo = gfile.GFile(tar_name, "w")
tf = TarFile(fileobj=tfo, mode='w')
for img_idx in xrange(NUM_IMGS):
img = data[0][img_idx, :, :, :]
imsave = Image.fromarray(img)
prefix = "CORRECT" if data[1][0, img_idx] == preds_thresh[
img_idx, PRED_IDX] else "FALSE_POS" if preds_thresh[
img_idx, PRED_IDX] == 1 else "FALSE_NEG"
file_name = "%s_%.2f_%d_%05d_%d.png" % (prefix, preds[
img_idx, PRED_IDX], batch, img_idx, data[1][0, img_idx])
# gf = gfile.GFile(file_name, "w")
file_string = StringIO()
imsave.save(file_string, "PNG")
tarinf = TarInfo(os.path.join(dir_name, file_name))
tarinf.size = file_string.tell()
file_string.seek(0)
tf.addfile(tarinf, file_string)
tf.close()
tfo.close()
# gf.close()
print "Wrote %d prediction PNGs to %s" % (preds.shape[0], tar_name)
else:
fig = pl.figure(3, figsize=(12, 9))
fig.text(
.4, .95, '%s test samples' %
('Mistaken' if self.only_errors else 'Random'))
if self.only_errors:
# what the net got wrong
if NUM_OUTPUTS > 1:
err_idx = [
i for i, p in enumerate(preds.argmax(axis=1))
if p not in n.where(data[2][:, i] > 0)[0]
]
else:
err_idx = n.where(data[1][0, :] != preds[:, 0].T)[0]
print err_idx
err_idx = r.sample(err_idx, min(len(err_idx), NUM_IMGS))
data[0], data[1], preds = data[0][:, err_idx], data[
1][:, err_idx], preds[err_idx, :]
import matplotlib.gridspec as gridspec
import matplotlib.colors as colors
cconv = colors.ColorConverter()
gs = gridspec.GridSpec(NUM_ROWS * 2,
NUM_COLS,
width_ratios=[1] * NUM_COLS,
height_ratios=[2, 1] * NUM_ROWS)
#print data[1]
for row in xrange(NUM_ROWS):
for col in xrange(NUM_COLS):
img_idx = row * NUM_COLS + col
if data[0].shape[0] <= img_idx:
break
pl.subplot(gs[(row * 2) * NUM_COLS + col])
#pl.subplot(NUM_ROWS*2, NUM_COLS, row * 2 * NUM_COLS + col + 1)
pl.xticks([])
pl.yticks([])
img = data[0][img_idx, :, :, :]
pl.imshow(img, interpolation='lanczos')
show_title = data[1].shape[0] == 1
true_label = [int(data[1][0, img_idx])
] if show_title else n.where(
data[1][:, img_idx] == 1)[0]
#print true_label
#print preds[img_idx,:].shape
#print preds[img_idx,:].max()
true_label_names = [label_names[i] for i in true_label]
img_labels = sorted(zip(preds[img_idx, :], label_names),
key=lambda x: x[0])[-NUM_TOP_CLASSES:]
#print img_labels
axes = pl.subplot(gs[(row * 2 + 1) * NUM_COLS + col])
height = 0.5
ylocs = n.array(range(NUM_TOP_CLASSES)) * height
pl.barh(ylocs, [l[0] for l in img_labels], height=height, \
color=['#ffaaaa' if l[1] in true_label_names else '#aaaaff' for l in img_labels])
#pl.title(", ".join(true_labels))
if show_title:
pl.title(", ".join(true_label_names),
fontsize=15,
fontweight='bold')
else:
print true_label_names
pl.yticks(ylocs + height / 2, [l[1] for l in img_labels],
x=1,
backgroundcolor=cconv.to_rgba('0.65', alpha=0.5),
weight='bold')
for line in enumerate(axes.get_yticklines()):
line[1].set_visible(False)
#pl.xticks([width], [''])
#pl.yticks([])
pl.xticks([])
pl.ylim(0, ylocs[-1] + height)
pl.xlim(0, 1)
avg_groups = ['warp','nowarp']
navg_groups = len(avg_groups)
if not donames:
load_convnets = sys.argv[1:] # just the paths
else:
load_convnets = sys.argv[1::2]; legend_names = sys.argv[2::2] # the paths and a shorter legend name
print 'Loading convnets ' + ', '.join(load_convnets)
for cost_idx in range(ncosts):
pl.figure(cost_idx + basefig)
if navg_groups > 1: allte = None; alltr = None
for cv,i in zip(load_convnets, range(len(load_convnets))):
cp = IGPUModel.load_checkpoint(cv)
numbatches, train_errors, test_errors = get_errors(cp, show_cost,cost_idx, norm=cost_idx == 0,
smooth=smooth, interp_test=dointerp)
#plot_errors(numbatches, train_errors, test_errors)
tr = np.array(train_errors, dtype=np.double); te = np.array(test_errors, dtype=np.double)
if navg_groups > 1:
if allte is None:
allte = np.zeros(list(te.shape)+[len(load_convnets)], dtype=np.double)
alltr = np.zeros(list(tr.shape)+[len(load_convnets)], dtype=np.double)
allte[:,i] = te; alltr[:,i] = tr
else:
if dolog: tr = np.log10(tr); te = np.log10(te)
pl.plot(tr, te, label=os.path.basename(legend_names[i] if donames else os.path.normpath(cv)))
if navg_groups > 1:
groups = np.arange(len(load_convnets),dtype=np.int64)
groups = groups.reshape([navg_groups,len(load_convnets)//navg_groups])
def plot_restored_img(self):
epoch, batch, data = self.get_next_batch(train=False)
NUM_ROWS = 8
NUM_COLS = 4
NUM_PATS = 3
NUM_IMGS = NUM_ROWS * NUM_COLS if not self.save_preds else data[
0].shape[1]
NUM_OUTPUTS = self.model_state['layers'][self.output_name]['outputs']
preds = n.zeros((NUM_IMGS, NUM_OUTPUTS), dtype=n.single)
# rand_idx = nr.randint(0, data[0].shape[1], NUM_IMGS)
rand_idx = range(NUM_IMGS)
if NUM_IMGS < data[0].shape[1]:
data = [n.require(d[:, rand_idx], requirements='C') for d in data]
print[d.shape for d in data], preds.shape
self.libmodel.startFeatureWriter(data, [preds], [self.output_name])
cost_outputs = IGPUModel.finish_batch(self)
costs, num_cases = cost_outputs[0], cost_outputs[1]
self.print_costs(cost_outputs)
# print preds
# data[0] = self.test_data_provider.get_plottable_data(data[0])
# preds = self.test_data_provider.get_plottable_data(preds.T)
patch_size = int(n.sqrt(data[0].shape[0]))
noisy_size = int(n.sqrt(data[1].shape[0]))
if self.save_preds:
pass
else:
fig = pl.figure(3, figsize=(12, 9))
fig.text(.4, .95, '%s %d test samples' % ('First', NUM_IMGS))
import matplotlib.gridspec as gridspec
import matplotlib.colors as colors
cconv = colors.ColorConverter()
gs = gridspec.GridSpec(NUM_ROWS,
NUM_COLS * NUM_PATS,
width_ratios=[1] * NUM_COLS * NUM_PATS,
height_ratios=[1] * NUM_ROWS)
#print data[1]
for row in xrange(NUM_ROWS):
for col in xrange(NUM_COLS):
img_idx = row * NUM_COLS + col
if data[0].shape[0] <= img_idx:
break
pl.subplot(gs[(row * NUM_COLS + col) * NUM_PATS])
#pl.subplot(NUM_ROWS*2, NUM_COLS, row * 2 * NUM_COLS + col + 1)
pl.xticks([])
pl.yticks([])
img = data[0][:, img_idx]
pl.imshow(self.plottable_data(img, patch_size, patch_size),
interpolation='nearest')
pl.subplot(gs[(row * NUM_COLS + col) * NUM_PATS + 1])
pl.xticks([])
pl.yticks([])
r_img = preds[img_idx, :]
mse = r_img - img
mse = mse * mse
mse = mse.mean()
psnr = -10 * n.log10(mse)
pl.title(('%.4f' % psnr), fontsize=10)
pl.imshow(self.plottable_data(r_img, patch_size,
patch_size),
interpolation='nearest')
pl.subplot(gs[(row * NUM_COLS + col) * NUM_PATS + 2])
pl.xticks([])
pl.yticks([])
n_img = data[1][:, img_idx]
border = (noisy_size - patch_size) / 2
nc_img = n_img.reshape(noisy_size,
noisy_size)[border:-border,
border:-border]
nc_img = nc_img.reshape(img.shape)
n_mse = nc_img - img
n_mse = n_mse * n_mse
n_mse = n_mse.mean()
n_psnr = -10 * n.log10(n_mse)
pl.title(('%.4f' % n_psnr), fontsize=10)
pl.imshow(self.plottable_data(n_img, noisy_size),
interpolation='nearest')
pl.savefig('restored.png')
def plot_predictions(self):
epoch, batch, data = self.get_next_batch(train=False) # get a test batch
num_classes = self.test_data_provider.get_num_classes()
NUM_ROWS = 2
NUM_COLS = 4
NUM_IMGS = NUM_ROWS * NUM_COLS if not self.save_preds else data[0].shape[1]
NUM_TOP_CLASSES = min(num_classes, 5) # show this many top labels
NUM_OUTPUTS = self.model_state['layers'][self.softmax_name]['outputs']
PRED_IDX = 1
if self.save_preds:
# print preds
if not os.path.exists(self.save_preds):
os.makedirs(self.save_preds)
# we process all the batches
while True:
# some constant
NUM_IMGS = data[0].shape[1]
NUM_TOP_CLASSES = min(num_classes, 5) # show this many top labels
NUM_OUTPUTS = self.model_state['layers'][self.softmax_name]['outputs']
PRED_IDX = 1
preds = n.zeros((NUM_IMGS, NUM_OUTPUTS), dtype=n.single)
# we only save the prediction result instead of the image
dir_name = 'predictions_batch_%d' % (batch)
tar_name = os.path.join(self.save_preds, dir_name)
# Run the model
print [d.shape for d in data], preds.shape
self.libmodel.startFeatureWriter(data, [preds], [self.softmax_name])
# in the mean while, prepare to load the next batch of data
new_epoch, batch, new_data = self.get_next_batch(train=False)
IGPUModel.finish_batch(self)
# swap the data
# concatenate the pred into the groud true label
preds=n.concatenate((n.transpose(data[1]), preds), axis=1);
tfo = open(tar_name, "wb");
cPickle.dump(preds, tfo)
tfo.close()
print "Wrote %d prediction PNGs to %s" % (preds.shape[0], tar_name)
if new_epoch!=epoch:
print "All batches process"
break;
data=new_data
else:
label_names = [lab.split(',')[0] for lab in self.test_data_provider.batch_meta['label_names']]
if self.only_errors:
preds = n.zeros((data[0].shape[1], NUM_OUTPUTS), dtype=n.single)
else:
preds = n.zeros((NUM_IMGS, NUM_OUTPUTS), dtype=n.single)
rand_idx = nr.randint(0, data[0].shape[1], NUM_IMGS)
if NUM_IMGS < data[0].shape[1]:
data = [n.require(d[:,rand_idx], requirements='C') for d in data]
# Run the model
print [d.shape for d in data], preds.shape
self.libmodel.startFeatureWriter(data, [preds], [self.softmax_name])
IGPUModel.finish_batch(self)
# print preds
data[0] = self.test_data_provider.get_plottable_data(data[0])
fig = pl.figure(3, figsize=(12,9))
fig.text(.4, .95, '%s test samples' % ('Mistaken' if self.only_errors else 'Random'))
if self.only_errors:
# what the net got wrong
if NUM_OUTPUTS > 1:
err_idx = [i for i,p in enumerate(preds.argmax(axis=1)) if p not in n.where(data[2][:,i] > 0)[0]]
else:
err_idx = n.where(data[1][0,:] != preds[:,0].T)[0]
print err_idx
err_idx = r.sample(err_idx, min(len(err_idx), NUM_IMGS))
data[0], data[1], preds = data[0][:,err_idx], data[1][:,err_idx], preds[err_idx,:]
import matplotlib.gridspec as gridspec
import matplotlib.colors as colors
cconv = colors.ColorConverter()
gs = gridspec.GridSpec(NUM_ROWS*2, NUM_COLS,
width_ratios=[1]*NUM_COLS, height_ratios=[2,1]*NUM_ROWS )
#print data[1]
for row in xrange(NUM_ROWS):
for col in xrange(NUM_COLS):
img_idx = row * NUM_COLS + col
if data[0].shape[0] <= img_idx:
break
pl.subplot(gs[(row * 2) * NUM_COLS + col])
#pl.subplot(NUM_ROWS*2, NUM_COLS, row * 2 * NUM_COLS + col + 1)
pl.xticks([])
pl.yticks([])
img = data[0][img_idx,:,:,:]
pl.imshow(img, interpolation='lanczos')
show_title = data[1].shape[0] == 1
true_label = [int(data[1][0,img_idx])] if show_title else n.where(data[1][:,img_idx]==1)[0]
#print true_label
#print preds[img_idx,:].shape
#print preds[img_idx,:].max()
true_label_names = [label_names[i] for i in true_label]
img_labels = sorted(zip(preds[img_idx,:], label_names), key=lambda x: x[0])[-NUM_TOP_CLASSES:]
#print img_labels
axes = pl.subplot(gs[(row * 2 + 1) * NUM_COLS + col])
height = 0.5
ylocs = n.array(range(NUM_TOP_CLASSES))*height
pl.barh(ylocs, [l[0] for l in img_labels], height=height, \
color=['#ffaaaa' if l[1] in true_label_names else '#aaaaff' for l in img_labels])
#pl.title(", ".join(true_labels))
if show_title:
pl.title(", ".join(true_label_names), fontsize=15, fontweight='bold')
else:
print true_label_names
pl.yticks(ylocs + height/2, [l[1] for l in img_labels], x=1, backgroundcolor=cconv.to_rgba('0.65', alpha=0.5), weight='bold')
for line in enumerate(axes.get_yticklines()):
line[1].set_visible(False)
#pl.xticks([width], [''])
#pl.yticks([])
pl.xticks([])
pl.ylim(0, ylocs[-1] + height)
pl.xlim(0, 1)