class IFaceRecognize():
def __init__(self,model_dir,gpu_id):
model_dir += '/' if not model_dir.endswith('/') else model_dir
assert(os.path.exists(model_dir+'facerecognize'))
prototxt = model_dir+"facerecognize/senet50_ft.prototxt"
weightfile = model_dir+'facerecognize/senet50_ft.caffemodel'
self.__gpu_id = gpu_id
self.__classifier = ObjTypeClassifier(prototxt,weightfile,None,gpu_id,[91.4953,103.8827,131.0912])
def extractFeature(self,im):
feature = self.__classifier.extractFeature(im,"classifier")
im_1 = np.flip(im,axis=1)
feature_1 = self.__classifier.extractFeature(im_1,"classifier")
result = np.hstack([feature,feature_1])
return result
def recognize(self,feature,im):
feature_1 = self.extractFeature(im)
sum_0 = np.dot(feature,feature)
sum_1 = np.dot(feature_1,feature_1)
result = np.dot(feature,feature_1)/(np.sqrt(sum_0)*np.sqrt(sum_1))
return result
def __init__(self,model_dir,gpu_id):
model_dir += '/' if not model_dir.endswith('/') else model_dir
assert(os.path.exists(model_dir+'facerecognize'))
prototxt = model_dir+"facerecognize/senet50_ft.prototxt"
weightfile = model_dir+'facerecognize/senet50_ft.caffemodel'
self.__gpu_id = gpu_id
self.__classifier = ObjTypeClassifier(prototxt,weightfile,None,gpu_id,[91.4953,103.8827,131.0912])
def __init__(self, model_dir, gpu_id=0):
deploy_file = os.path.join(model_dir,
"patentRetrieval/deploy.prototxt")
assert os.path.exists(deploy_file)
weights_file = os.path.join(model_dir,
"patentRetrieval/final.caffemodel")
assert os.path.exists(weights_file)
self.__extracter = ObjTypeClassifier(deploy_file, weights_file, gpu_id)
self.__feature_len = 1024
class IVehicleColorClassifier():
def __init__(self, modelDir, gpu_id):
modelDir += '/' if not modelDir.endswith('/') else modelDir
assert (os.path.exists(modelDir + 'vehicleTypeColor'))
prototxt = modelDir + "vehicleTypeColor/vehicleTypeColorD10deploy.prototxt"
assert (prototxt)
weightfile = modelDir + 'vehicleTypeColor/vehicleTypeColorD10model.dat'
assert (weightfile)
meanfile = modelDir + 'vehicleTypeColor/vehicleTypeColorD10mean.dat'
assert (meanfile)
labelfile = modelDir + 'vehicleTypeColor/vehicleTypeColorID10.txt'
assert (labelfile)
self.__gpu_id = gpu_id
self.__classifier = ObjTypeClassifier(prototxt, weightfile, meanfile,
gpu_id)
self.__labelMap = [
line.strip() for line in open(labelfile).readlines()
]
def classify(self, im):
result = {}
info = self.__classifier.classify(im, 'fc7')
result['category'] = self.__labelMap[info[0]]
result['score'] = info[1]
return result
class IPatentTypeClassifier():
def __init__(self, modelDir, gpu_id):
modelDir += '/' if not modelDir.endswith('/') else modelDir
assert (os.path.exists(modelDir + 'patentClass'))
prototxt = modelDir + "patentClass/patentimageclass.prototxt"
assert (prototxt)
weightfile = modelDir + 'patentClass/bvlc_googlenet_iter_14000.caffemodel'
assert (weightfile)
meanfile = modelDir + 'patentClass/patentimageclassmean.binaryproto'
assert (meanfile)
labelfile = modelDir + 'patentClass/patentAllTypeID4.txt'
assert (labelfile)
self.__gpu_id = gpu_id
self.__classifier = ObjTypeClassifier(prototxt, weightfile, meanfile,
gpu_id)
self.__labelMap = [
line.strip() for line in open(labelfile).readlines()
]
def classify(self, im):
result = {}
info = self.__classifier.classify(im, 'loss2/fc')
result['category'] = self.__labelMap[info[0]]
result['score'] = info[1]
return result
def __init__(self,modelDir,gpu_id):
modelDir += '/' if not modelDir.endswith('/') else modelDir
assert(os.path.exists(modelDir+'vehicleTypeTail'))
prototxt = modelDir+"vehicleTypeTail/vehicleTypeTailD0deploy.prototxt"
assert(prototxt)
weightfile = modelDir+'vehicleTypeTail/vehicleTypeTailD0model.dat'
assert(weightfile)
meanfile = modelDir+'vehicleTypeTail/vehicleTypeTailD0mean.dat'
assert(meanfile)
labelfile = modelDir+'vehicleTypeTail/vehicleTypeTailID4.txt'
assert(labelfile)
self.__gpu_id = gpu_id
self.__classifier = ObjTypeClassifier(prototxt,weightfile,meanfile,gpu_id)
self.__labelMap = [line.strip() for line in open(labelfile).readlines()]
def __init__(self, modelDir, gpu_id):
modelDir += '/' if not modelDir.endswith('/') else modelDir
assert (os.path.exists(modelDir + 'patentClass'))
prototxt = modelDir + "patentClass/patentimageclass.prototxt"
assert (prototxt)
weightfile = modelDir + 'patentClass/bvlc_googlenet_iter_14000.caffemodel'
assert (weightfile)
meanfile = modelDir + 'patentClass/patentimageclassmean.binaryproto'
assert (meanfile)
labelfile = modelDir + 'patentClass/patentAllTypeID4.txt'
assert (labelfile)
self.__gpu_id = gpu_id
self.__classifier = ObjTypeClassifier(prototxt, weightfile, meanfile,
gpu_id)
self.__labelMap = [
line.strip() for line in open(labelfile).readlines()
]
def __init__(self, model_dir, gpu_id=0):
deploy_file = os.path.join(model_dir,
"faceretrieval/resnet50_ft.prototxt")
assert os.path.exists(deploy_file)
weights_file = os.path.join(model_dir,
"faceretrieval/resnet50_ft.caffemodel")
assert os.path.exists(weights_file)
xml_file = os.path.join(model_dir, "faceretrieval/config95-V.xml")
assert os.path.exists(xml_file)
fs = cv2.FileStorage(xml_file, cv2.FILE_STORAGE_READ)
self.__pca_mean = fs.getNode("mean").mat()
self.__pca_eigen_vector = fs.getNode("eigen_vector").mat()
self.__feature_len = self.__pca_eigen_vector.shape[0]
mean_value = [91.4953, 103.8827, 131.0912]
self.__extracter = ObjTypeClassifier(deploy_file,
weights_file,
meanvalue=mean_value)
class PatentRetrieval:
def __init__(self, model_dir, gpu_id=0):
deploy_file = os.path.join(model_dir,
"patentRetrieval/deploy.prototxt")
assert os.path.exists(deploy_file)
weights_file = os.path.join(model_dir,
"patentRetrieval/final.caffemodel")
assert os.path.exists(weights_file)
self.__extracter = ObjTypeClassifier(deploy_file, weights_file, gpu_id)
self.__feature_len = 1024
def create_retrieval_model(self, retrieval_model_path):
assert (os.path.exists(retrieval_model_path))
retrieval_model = faiss.read_index(retrieval_model_path)
return retrieval_model
def extractFeature(self, im):
try:
feature = self.__extracter.extractFeature(im, 'loss2/fc')
except:
return []
return feature
def buildRetrievalDatabase(self, features, retrieval_model_path):
if os.path.exists(retrieval_model_path):
retrieval_model = faiss.read_index(retrieval_model_path)
else:
retrieval_model = faiss.IndexFlatL2(self.__feature_len)
features = np.array(features)
retrieval_model.add(features)
faiss.write_index(retrieval_model, retrieval_model_path)
return retrieval_model
def query(self, im, k, retrieval_model):
feature = self.extractFeature(im)
feature = np.array([feature])
res = retrieval_model.search(feature, k)
return res[1][0], res[0][0]
class FaceRetrieval:
def __init__(self, model_dir, gpu_id=0):
deploy_file = os.path.join(model_dir,
"faceretrieval/resnet50_ft.prototxt")
assert os.path.exists(deploy_file)
weights_file = os.path.join(model_dir,
"faceretrieval/resnet50_ft.caffemodel")
assert os.path.exists(weights_file)
xml_file = os.path.join(model_dir, "faceretrieval/config95-V.xml")
assert os.path.exists(xml_file)
fs = cv2.FileStorage(xml_file, cv2.FILE_STORAGE_READ)
self.__pca_mean = fs.getNode("mean").mat()
self.__pca_eigen_vector = fs.getNode("eigen_vector").mat()
self.__feature_len = self.__pca_eigen_vector.shape[0]
mean_value = [91.4953, 103.8827, 131.0912]
self.__extracter = ObjTypeClassifier(deploy_file,
weights_file,
meanvalue=mean_value)
def create_retrieval_model(self, retrieval_model_path):
assert (os.path.exists(retrieval_model_path))
retrieval_model = AnnoyIndex(self.__feature_len)
retrieval_model.load(retrieval_model_path)
return retrieval_model
def extractFeature(self, im):
try:
feature = self.__extracter.extractFeature(im, "classifier")
im_1 = np.flip(im, axis=1)
feature_1 = self.__extracter.extractFeature(im_1, "classifier")
result = np.hstack([feature, feature_1])
result = result.reshape(1, -1)
res = cv2.PCAProject(result, self.__pca_mean,
self.__pca_eigen_vector)
return res.reshape(-1).tolist()
except:
return []
def buildRetrievalDatabase(self, features, retrieval_model_path):
self.__database = AnnoyIndex(self.__feature_len)
index = 0
for feature in features:
self.__database.add_item(index, feature)
index += 1
self.__database.build(2 * self.__feature_len)
self.__database.save(retrieval_model_path)
def query(self, im, k, retrieval_model):
feature = self.extractFeature(im)
res = retrieval_model.get_nns_by_vector(feature,
k,
include_distances=False)
return res