img[:, :, 0] = cv2.equalizeHist(img[:, :, 0])
img[:, :, 1] = cv2.equalizeHist(img[:, :, 1])
img[:, :, 2] = cv2.equalizeHist(img[:, :, 2])
#Image Resizing
img = cv2.resize(img, (img_width, img_height),
interpolation=cv2.INTER_CUBIC)
return img
'''
Reading mean image, caffe model and its weights
'''
#Read mean image
mean_blob = caffe_pb2.BlobProto()
#with open('/home/ubuntu/deeplearning-cats-dogs-tutorial/input/mean.binaryproto') as f:
#with open('/home/s2c/pkg/local/caffe-master_cuDNN/data/ilsvrc12/imagenet_mean.binaryproto.bk', 'rb') as f:
with open('./imagenet_mean.binaryproto', 'rb') as f:
mean_blob.ParseFromString(f.read())
mean_array = np.asarray(mean_blob.data, dtype=np.float32).reshape(
(mean_blob.channels, mean_blob.height, mean_blob.width))
print('mean_array type = ', type(mean_array))
print('mean array shape = ', mean_array.shape)
print('mean_blob channels = ', mean_blob.channels)
print('mean_blob height = ', mean_blob.height)
print('mean_blob width = ', mean_blob.width)
#Read model architecture and trained model's weights
net = caffe.Net('./deploy.prototxt', './bvlc_alexnet.caffemodel', caffe.TEST)
def Predict(self):
colours = {
1: 'black',
2: 'blue',
3: 'brown',
4: 'darkblue',
5: 'gold',
6: 'green',
7: 'maroon',
8: 'orange',
9: 'red',
10: 'silver',
11: 'white',
12: 'yellow'
}
types = {'suv': 1, 'hatchback': 2, 'sedan': 3, 'carrier': 4, 'van': 5}
mean_blob = caffe_pb2.BlobProto()
with open('input/mean.binaryproto') as f:
mean_blob.ParseFromString(f.read())
mean_array = np.asarray(mean_blob.data, dtype=np.float32).reshape(
(mean_blob.channels, mean_blob.height, mean_blob.width))
net = caffe.Net('deploy.prototxt', 'caffe_model_iter_100.caffemodel',
caffe.TEST)
net2 = caffe.Net('deploy2.prototxt', 'caffe_type_iter_100.caffemodel',
caffe.TEST)
transformer = caffe.io.Transformer(
{'data': net.blobs['data'].data.shape})
#transformer.set_mean('data', mean_array)
transformer.set_transpose('data', (2, 0, 1))
test_img_paths = [str(self.filename)]
test_ids = []
preds = []
preds2 = []
for img_path in test_img_paths:
img = cv2.imread(img_path, cv2.IMREAD_COLOR)
img = self.transform_img(img,
img_width=IMAGE_WIDTH,
img_height=IMAGE_HEIGHT)
net.blobs['data'].data[...] = transformer.preprocess('data', img)
net2.blobs['data'].data[...] = transformer.preprocess('data', img)
out = net.forward(data=np.asarray(net.blobs['data'].data[...]))
out2 = net2.forward(data=np.asarray(net.blobs['data'].data[...]))
pred_probas = out['prob']
pred_probas2 = out2['prob']
highest = pred_probas[0][0]
highest2 = pred_probas2[0][0]
index = 0
index2 = 0
test_ids = test_ids + [img_path]
result_string = ""
for x in range(len(colours)):
if pred_probas[0][x] > highest:
highest = pred_probas[0][x]
index = x
for x in range(len(types)):
if pred_probas2[0][x] > highest2:
highest2 = pred_probas[0][x]
index2 = x
print img_path
print colours[index + 1]
print types[index2]
print '-------'
result_string = colours[index + 1] + " " + types[index2 + 1]
with open("results.csv", "w") as f:
f.write("id,label\n")
for i in range(len(test_ids)):
f.write(str(test_ids[i]) + "," + str(preds[i]) + "\n")
self.ui.result.setText(result_string)
f.close()
def TranslateModel(
cls,
caffe_net,
pretrained_net,
is_test=False,
input_mean=None,
net_state=None,
):
net_state = caffe_pb2.NetState() if net_state is None else net_state
net = caffe2_pb2.NetDef()
net.name = caffe_net.name
net_params = caffe2_pb2.TensorProtos()
if len(caffe_net.layer) == 0:
raise ValueError(
'I think something is wrong. This translation script '
'only accepts new style layers that are stored in the '
'layer field.')
if input_mean:
caffenet_mean = caffe_pb2.BlobProto()
caffenet_mean.ParseFromString(open(input_mean, 'rb').read())
mean_ = utils.CaffeBlobToNumpyArray(caffenet_mean)
mean_tensor = utils.NumpyArrayToCaffe2Tensor(mean_, 'mean_')
net_params.protos.extend([mean_tensor])
mean_op = caffe2_pb2.OperatorDef()
mean_op.type = 'Sub'
mean_op.input.extend(['data_', 'mean_'])
# Assume that input blob's name is "data"
mean_op.output.extend(['data'])
net.op.extend([mean_op])
i = 0
while i < len(caffe_net.layer):
if not _ShouldInclude(net_state, caffe_net.layer[i]):
log.info('Current net state does not need layer {}'.format(
caffe_net.layer[i].name))
continue
log.info('Translate layer {}'.format(caffe_net.layer[i].name))
# Get pretrained one
pretrained_layers_index = ([
l for l in xrange(len(pretrained_net.layer))
if pretrained_net.layer[l].name == caffe_net.layer[i].name
] + [
l for l in xrange(len(pretrained_net.layers))
if pretrained_net.layers[l].name == caffe_net.layer[i].name
])
is_bn = False
if len(pretrained_layers_index) > 1:
raise ValueError(
'huh? more than one pretrained layer of one name?')
elif len(pretrained_layers_index) == 1:
if pretrained_net.layer[
pretrained_layers_index[0]].type == "BatchNorm":
# A Scale layer should follow BatchNorm layer
# according to paper https://arxiv.org/abs/1502.03167.
assert pretrained_net.layer[pretrained_layers_index[0] +
1].type == "Scale"
pretrained_blobs = [utils.CaffeBlobToNumpyArray(blob)
for blob in pretrained_net.layer[pretrained_layers_index[0]].blobs] + \
[utils.CaffeBlobToNumpyArray(blob)
for blob in pretrained_net.layer[pretrained_layers_index[0] + 1].blobs]
is_bn = True
else:
pretrained_blobs = [
utils.CaffeBlobToNumpyArray(blob) for blob in
pretrained_net.layer[pretrained_layers_index[0]].blobs
]
else:
# No pretrained layer for the given layer name. We'll just pass
# no parameter blobs.
# print 'No pretrained layer for layer', layer.name
pretrained_blobs = []
operators, params = cls.TranslateLayer(caffe_net.layer[i],
pretrained_blobs, is_test)
net.op.extend(operators)
net_params.protos.extend(params)
if is_bn:
i += 2
else:
i += 1
return net, net_params
def load_weights(self, caffemodel):
if self.has_mean:
print('mean_file', self.mean_file)
mean_blob = caffe_pb2.BlobProto()
mean_blob.ParseFromString(open(self.mean_file, 'rb').read())
if 'input_shape' in self.net_info['props']:
channels = int(self.net_info['props']['input_shape']['dim'][1])
height = int(self.net_info['props']['input_shape']['dim'][2])
width = int(self.net_info['props']['input_shape']['dim'][3])
else:
channels = int(self.net_info['props']['input_dim'][1])
height = int(self.net_info['props']['input_dim'][2])
width = int(self.net_info['props']['input_dim'][3])
mu = np.array(mean_blob.data)
mu.resize(channels, height, width)
mu = mu.mean(1).mean(1)
mean_img = torch.from_numpy(mu).view(channels, 1, 1).expand(
channels, height, width).float()
self.register_buffer('mean_img',
torch.zeros(channels, height, width))
self.mean_img.copy_(mean_img)
model = parse_caffemodel(caffemodel)
layers = model.layer
if len(layers) == 0:
print('Using V1LayerParameter')
layers = model.layers
lmap = {}
for l in layers:
lmap[l.name] = l
layers = self.net_info['layers']
layer_num = len(layers)
i = 0
while i < layer_num:
layer = layers[i]
lname = layer['name']
if 'include' in layer and 'phase' in layer['include']:
phase = layer['include']['phase']
lname = lname + '.' + phase
ltype = layer['type']
# if not lmap.has_key(lname):
if lname not in lmap:
i = i + 1
continue
if ltype in ['Convolution', 'Deconvolution']:
print('load weights %s' % lname)
convolution_param = layer['convolution_param']
bias = True
# if convolution_param.has_key('bias_term') and convolution_param['bias_term'] == 'false':
if 'bias_term' in convolution_param and convolution_param[
'bias_term'] == 'false':
bias = False
#weight_blob = lmap[lname].blobs[0]
#print('caffe weight shape', weight_blob.num, weight_blob.channels, weight_blob.height, weight_blob.width)
caffe_weight = np.array(lmap[lname].blobs[0].data)
caffe_weight = torch.from_numpy(caffe_weight).view_as(
self.models[lname].weight)
self.models[lname].weight.data.copy_(caffe_weight)
if bias and len(lmap[lname].blobs) > 1:
self.models[lname].bias.data.copy_(
torch.from_numpy(np.array(lmap[lname].blobs[1].data)))
#print("convlution %s has bias" % lname)
i = i + 1
elif ltype == 'BatchNorm':
print('load weights %s' % lname)
self.models[lname].running_mean.copy_(
torch.from_numpy(
np.array(lmap[lname].blobs[0].data) /
lmap[lname].blobs[2].data[0]))
self.models[lname].running_var.copy_(
torch.from_numpy(
np.array(lmap[lname].blobs[1].data) /
lmap[lname].blobs[2].data[0]))
i = i + 1
elif ltype == 'Scale':
print('load weights %s' % lname)
self.models[lname].weight.data.copy_(
torch.from_numpy(np.array(lmap[lname].blobs[0].data)))
self.models[lname].bias.data.copy_(
torch.from_numpy(np.array(lmap[lname].blobs[1].data)))
i = i + 1
elif ltype == 'Normalize':
print('load weights %s' % lname)
self.models[lname].weight.data.copy_(
torch.from_numpy(np.array(lmap[lname].blobs[0].data)))
i = i + 1
elif ltype == 'InnerProduct':
print('load weights %s' % lname)
if type(self.models[lname]) == nn.Sequential:
self.models[lname][1].weight.data.copy_(
torch.from_numpy(np.array(
lmap[lname].blobs[0].data)).view_as(
self.models[lname][1].weight))
if len(lmap[lname].blobs) > 1:
self.models[lname][1].bias.data.copy_(
torch.from_numpy(
np.array(lmap[lname].blobs[1].data)))
else:
self.models[lname].weight.data.copy_(
torch.from_numpy(np.array(lmap[lname].blobs[0].data)))
if len(lmap[lname].blobs) > 1:
self.models[lname].bias.data.copy_(
torch.from_numpy(
np.array(lmap[lname].blobs[1].data)))
i = i + 1
else:
if not ltype in SUPPORTED_LAYERS:
print('load_weights: unknown type %s' % ltype)
i = i + 1
def as_blob(array):
blob = pb2.BlobProto()
blob.shape.dim.extend(array.shape)
blob.data.extend(array.astype(float).flat)
return blob
def mean_blob_fn(self):
mean_blob = caffe_pb2.BlobProto()
with open('../mean.binaryproto') as f:
mean_blob.ParseFromString(f.read())
return mean_blob
def binartproto2npy(mean_file):
blob = caffe_pb2.BlobProto()
blob.ParseFromString(open(mean_file, 'rb').read())
arr = np.array(caffe.io.blobproto_to_array(blob))
out = arr[0]
return out
def load_mean_binary(mean_file):
blob = caffe_pb2.BlobProto()
bin_mean = open(mean_file, 'rb').read() #proto file
blob.ParseFromString(bin_mean)
arr_mean = caffe.io.blobproto_to_array(blob)
return arr_mean[0]
n = 49
dxdy.append([-abs(a1[0] - a2[0]) / n, abs(a1[1] - a2[1]) / n])
dxdy_2 = []
for i in range(1, len(pointers_lat_2) - 1):
direct = 2
p1 = label2coord(i - 1)
p2 = label2coord(i)
p3 = label2coord(i + 1)
a1 = [abs(p1[0] + p2[0]) / 2, abs(p1[1] + p2[1]) / 2]
a2 = [abs(p2[0] + p3[0]) / 2, abs(p2[1] + p3[1]) / 2]
n = 49
dxdy_2.append([abs(a1[0] - a2[0]) / n, -abs(a1[1] - a2[1]) / n])
#Read mean images, caffe model architectures and their trained weights
mean_blob = caffe_pb2.BlobProto()
with open('./input/SFMV/mean.binaryproto') as f:
mean_blob.ParseFromString(f.read())
mean_array = np.asarray(mean_blob.data, dtype=np.float32).reshape(
(mean_blob.channels, mean_blob.height, mean_blob.width))
net = caffe.Net('./caffe/SFMV/deploy.prototxt',
'./caffe/SFMV/googlenet_train_iter_186000.caffemodel',
caffe.TEST)
#Define image transformers
transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})
transformer.set_mean('data', mean_array)
transformer.set_transpose('data', (2, 0, 1))
#Read mean images, caffe model architectures and their trained weights
mean_blob_way = caffe_pb2.BlobProto()
def _load_binaryproto(file):
blob = caffe_pb2.BlobProto()
data = open(file, 'rb').read()
blob.ParseFromString(data)
arr = np.array(caffe.io.blobproto_to_array(blob))
return arr[0]
