def __init__(self, model_config):
# Load all of the labels into memory
labels_path = self._get_absolute_path('data/synset.txt')
with open(labels_path, 'r') as f:
# synset.txt has a weird format, just get the first label
self.labels = [
' '.join(l.rstrip().split()[1:]).split(',')[0] for l in f
]
self.Batch = namedtuple('Batch', ['data'])
# Load the ONNX model
model_path = self._get_absolute_path(model_config['model_path'])
sym, arg_params, aux_params = import_model(model_path)
# Determine and set context
if len(mx.test_utils.list_gpus()) == 0:
print('No GPUs available')
ctx = mx.cpu()
else:
print('Using GPU')
ctx = mx.gpu(0)
# Load module
self.mod = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
self.mod.bind(for_training=False,
data_shapes=[('data', (1, 3, 224, 224))],
label_shapes=self.mod._label_shapes)
self.mod.set_params(arg_params,
aux_params,
allow_missing=True,
allow_extra=True)
def get_model(ctx, model):
image_size = (112, 112)
# Import ONNX model
sym, arg_params, aux_params = import_model(model)
# Define and binds parameters to the network
model = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))])
model.set_params(arg_params, aux_params)
return model
def get_PR(model_path):
ctx = mx.gpu(0)
image_size = (112, 112)
sym, arg_params, aux_params = import_model(model_path + '/' +
'lib-pm.44.to.bin')
model = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))])
model.set_params(arg_params, aux_params)
return model
def get_recognition_model(ctx, model):
'''
Loads ONNX model into MXNet symbols and params,
defines model using symbol file and binds parameters to the model using params file.
'''
image_size = (112, 112)
# Import ONNX model
sym, arg_params, aux_params = import_model(model)
# Define and binds parameters to the network
model = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))])
model.set_params(arg_params, aux_params)
return model
def __init__(self):
# load config file
config = json.load(open("model/config.json"))
# get the image processor
self._imageProcessor = ImageProcessor(config)
# get context - cpu
ctx = mx.cpu()
image_size = (112,112)
# Import ONNX model
sym, arg_params, aux_params = import_model('model/model.onnx')
# Define and binds parameters to the network
model = mx.mod.Module(symbol=sym, context=ctx, label_names = None)
model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))])
model.set_params(arg_params, aux_params)
self._model = model
def __init__(self):
# load config file
config = json.load(open("model/config.json"))
# get the image processor
self._imageProcessor = ImageProcessor(config)
# get context - cpu
ctx = mx.cpu()
sym, arg_params, aux_params = import_model('model/model.onnx')
# Load module
mod = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
mod.bind(for_training=False,
data_shapes=[('data', (1, 3, 224, 224))],
label_shapes=mod._label_shapes)
mod.set_params(arg_params,
aux_params,
allow_missing=True,
allow_extra=True)
self._model = mod
def __init__(self, args):
self.args = args
print('NNResNet50v2 initializer called')
print('NNResNet50v2 model being loaded')
model_path= 'nn/onnx/resnet50v2.onnx'
sym, arg_params, aux_params = import_model(model_path)
if len(mx.test_utils.list_gpus())==0:
ctx = mx.cpu()
else:
ctx = mx.gpu(0)
self.Batch = namedtuple('Batch', ['data'])
self.img_path = 'nn/onnx/file.jpg'
self.mod = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
self.mod.bind(for_training=False, data_shapes=[('data', (1,3,224,224))],
label_shapes=self.mod._label_shapes)
self.mod.set_params(arg_params, aux_params, allow_missing=True, allow_extra=True)
with open('nn/onnx/synset.txt', 'r') as f:
self.labels = [l.rstrip() for l in f]
print('NNResNet50v2 model loading done')
print('NNTemplate initializer called')
import mxnet as mx
import matplotlib.pyplot as plt
import numpy as np
from collections import namedtuple
from mxnet.gluon.data.vision import transforms
from mxnet.contrib.onnx.onnx2mx.import_model import import_model
plt.switch_backend('Agg')
#mx.test_utils.download('https://s3.amazonaws.com/onnx-model-zoo/synset.txt')
with open('synset.txt', 'r') as f:
labels = [l.rstrip() for l in f]
# Enter path to the ONNX model file
model_path = 'resnet18-v2-7.onnx'
sym, arg_params, aux_params = import_model(model_path)
Batch = namedtuple('Batch', ['data'])
def get_image(path, show=False):
img = mx.image.imread(path)
if img is None:
return None
if show:
plt.imshow(img.asnumpy())
plt.axis('off')
return img
def preprocess(img):
def prt_model(model_id, output_layer):
# ResNet50
print('loading model...')
# 0 ResNet50
if model_id == 0:
if output_layer == 0: # Keras
model_name = 'ResNet50_Keras'
model = ResNet50(weights='imagenet', include_top=False)
img_size = 224
#img_size = 30
model_path = '../models/resnet50.onnx'
datain = 'gpu_0/data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'ResNet50_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'ResNet50_ONNX_P'
layerout = 'flatten0_output'
# 1 VGG16
elif model_id == 1:
if output_layer == 0: # Keras
model_name = 'VGG16_Keras'
model = VGG16(weights='imagenet', include_top=False)
#img_size = 224
img_size = 60
model_path = '../models/vgg16.onnx'
datain = 'data'
if output_layer == 1: # ONNX - Last Layer
model_name = 'VGG16_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'VGG16_ONNX_P'
layerout = 'flatten2_output'
# 2 VGG19
elif model_id == 2:
if output_layer == 0: # Keras
model_name = 'VGG19_Keras'
model = VGG19(weights='imagenet', include_top=False)
#img_size = 224
img_size = 60
model_path = '../models/vgg19.onnx'
datain = 'data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'VGG19_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'VGG19_ONNX_P'
layerout = 'flatten2_output'
# 3 InceptionV3
elif model_id == 3:
if output_layer == 0: # Keras
model_name = 'InceptionV3_Keras'
model = InceptionV3(weights='imagenet', include_top=False)
#img_size = 224
img_size = 90
# 4 InceptionResNetV2
#elif model_id == 4:
# if output_layer == 0: # Keras
# model_name = 'InceptionResNetV2_Keras'
# model = InceptionResNetV2(weights='imagenet', include_top=False)
# #img_size = 224
# img_size = 90
# 5 Xception
elif model_id == 5:
if output_layer == 0: # Keras
model_name = 'Xception_Keras'
model = Xception(weights='imagenet', include_top=False)
#img_size = 224
img_size = 30
# 6 MobileNet
elif model_id == 6:
model_path = '../models/mobilenet.onnx'
datain = 'data'
if output_layer == 1: # ONNX - Last Layer
model_name = 'MobileNet_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'MobileNet_ONNX_P'
layerout = 'mobilenetv20_features_pool0_fwd_output' # not good
# 7 SqueezeNet
elif model_id == 7:
model_path = '../models/squeezenet.onnx'
datain = 'data'
if output_layer == 1: # ONNX - Last Layer
model_name = 'SqueezeNet_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'SqueezeNet_ONNX_P'
layerout = 'pooling3_output' # not good
# 8 AlexNet
elif model_id == 8:
model_path = '../models/alexnet.onnx'
datain = 'data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'AlexNet_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'AlexNet_ONNX_P'
layerout = 'flatten2_output'
# 9 GoogleNet
elif model_id == 9:
model_path = '../models/googlenet.onnx'
datain = 'data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'GoogleNet_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'GoogleNet_ONNX_P'
layerout = 'flatten0_output' # not good
# 10 ShuffleNet
elif model_id == 10:
model_path = '../models/shufflenet.onnx'
datain = 'gpu_0/data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'ShuffleNet_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'ShuffleNet_ONNX_P'
layerout = 'flatten0_output' # not good
# 11 DenseNet121
elif model_id == 11:
model_path = '../models/densenet121.onnx'
datain = 'data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'DenseNet121_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'DenseNet121_ONNX_P'
layerout = 'pad124_output' # not good
# 12 ZfNet512
elif model_id == 12:
model_path = '../models/zfnet512.onnx'
datain = 'gpu_0/data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'ZfNet512_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'ZfNet512_ONNX_P'
layerout = 'flatten2_output' # not good
# 13 RCNN_ILSVRC13
elif model_id == 13:
model_path = '../models/rcnn_ilsvrc13.onnx'
datain = 'data_0'
if output_layer == 1: # ONNX - Last Layer
model_name = 'RCNN_ILSVRC13_ONNX_L'
layerout = 'last'
elif output_layer == 2: # ONNX - Previous Layer
model_name = 'RCNN_ILSVRC13_ONNX_P'
layerout = 'flatten2_output' # not good
if output_layer > 0:
print('loading model ' + model_path + '...')
sym, arg_params, aux_params = import_model(model_path)
if len(mx.test_utils.list_gpus()) == 0:
ctx = mx.cpu()
else:
ctx = mx.gpu(0)
all_layers = sym.get_internals()
print(all_layers.list_outputs())
if layerout == 'last':
sym3 = sym
else:
sym3 = all_layers[layerout]
model = mx.mod.Module(symbol=sym3,
context=ctx,
label_names=None,
data_names=[datain])
image_size = (224, 224)
img_size = image_size[1]
model.bind(data_shapes=[(datain, (1, 3, image_size[0],
image_size[1]))])
model.set_params(arg_params, aux_params)
print('model ' + model_name + ' loaded.')
return model, img_size, model_name
