OpenCV  4.7.0-dev
Open Source Computer Vision
Conversion of TensorFlow Classification Models and Launch with OpenCV Python
Original author Anastasia Murzova
Compatibility OpenCV >= 4.5

Goals

In this tutorial you will learn how to:

We will explore the above-listed points by the example of MobileNet architecture.

Introduction

Let's briefly view the key concepts involved in the pipeline of TensorFlow models transition with OpenCV API. The initial step in conversion of TensorFlow models into cv.dnn.Net is obtaining the frozen TF model graph. Frozen graph defines the combination of the model graph structure with kept values of the required variables, for example, weights. Usually the frozen graph is saved in protobuf (.pb) files. After the model .pb file was generated it can be read with cv.dnn.readNetFromTensorflow function.

Requirements

To be able to experiment with the below code you will need to install a set of libraries. We will use a virtual environment with python3.7+ for this:

virtualenv -p /usr/bin/python3.7 <env_dir_path>
source <env_dir_path>/bin/activate

For OpenCV-Python building from source, follow the corresponding instructions from the Introduction to OpenCV.

Before you start the installation of the libraries, you can customize the requirements.txt, excluding or including (for example, opencv-python) some dependencies. The below line initiates requirements installation into the previously activated virtual environment:

pip install -r requirements.txt

Practice

In this part we are going to cover the following points:

  1. create a TF classification model conversion pipeline and provide the inference
  2. evaluate and test TF classification models

If you'd like merely to run evaluation or test model pipelines, the "Model Conversion Pipeline" tutorial part can be skipped.

Model Conversion Pipeline

The code in this subchapter is located in the dnn_model_runner module and can be executed with the line:

python -m dnn_model_runner.dnn_conversion.tf.classification.py_to_py_mobilenet

The following code contains the description of the below-listed steps:

  1. instantiate TF model
  2. create TF frozen graph
  3. read TF frozen graph with OpenCV API
  4. prepare input data
  5. provide inference
# initialize TF MobileNet model
original_tf_model = MobileNet(
include_top=True,
weights="imagenet"
)
# get TF frozen graph path
full_pb_path = get_tf_model_proto(original_tf_model)
# read frozen graph with OpenCV API
opencv_net = cv2.dnn.readNetFromTensorflow(full_pb_path)
print("OpenCV model was successfully read. Model layers: \n", opencv_net.getLayerNames())
# get preprocessed image
input_img = get_preprocessed_img("../data/squirrel_cls.jpg")
# get ImageNet labels
imagenet_labels = get_imagenet_labels("../data/dnn/classification_classes_ILSVRC2012.txt")
# obtain OpenCV DNN predictions
get_opencv_dnn_prediction(opencv_net, input_img, imagenet_labels)
# obtain TF model predictions
get_tf_dnn_prediction(original_tf_model, input_img, imagenet_labels)

To provide model inference we will use the below squirrel photo (under CC0 license) corresponding to ImageNet class ID 335:

fox squirrel, eastern fox squirrel, Sciurus niger
squirrel_cls.jpg
Classification model input image

For the label decoding of the obtained prediction, we also need imagenet_classes.txt file, which contains the full list of the ImageNet classes.

Let's go deeper into each step by the example of pretrained TF MobileNet:

# initialize TF MobileNet model
original_tf_model = MobileNet(
include_top=True,
weights="imagenet"
)
# define the directory for .pb model
pb_model_path = "models"
# define the name of .pb model
pb_model_name = "mobilenet.pb"
# create directory for further converted model
os.makedirs(pb_model_path, exist_ok=True)
# get model TF graph
tf_model_graph = tf.function(lambda x: tf_model(x))
# get concrete function
tf_model_graph = tf_model_graph.get_concrete_function(
tf.TensorSpec(tf_model.inputs[0].shape, tf_model.inputs[0].dtype))
# obtain frozen concrete function
frozen_tf_func = convert_variables_to_constants_v2(tf_model_graph)
# get frozen graph
frozen_tf_func.graph.as_graph_def()
# save full tf model
tf.io.write_graph(graph_or_graph_def=frozen_tf_func.graph,
logdir=pb_model_path,
name=pb_model_name,
as_text=False)

After the successful execution of the above code, we will get a frozen graph in models/mobilenet.pb.

# get TF frozen graph path
full_pb_path = get_tf_model_proto(original_tf_model)
# read the image
input_img = cv2.imread(img_path, cv2.IMREAD_COLOR)
input_img = input_img.astype(np.float32)
# define preprocess parameters
mean = np.array([1.0, 1.0, 1.0]) * 127.5
scale = 1 / 127.5
# prepare input blob to fit the model input:
# 1. subtract mean
# 2. scale to set pixel values from 0 to 1
input_blob = cv2.dnn.blobFromImage(
image=input_img,
scalefactor=scale,
size=(224, 224), # img target size
mean=mean,
swapRB=True, # BGR -> RGB
crop=True # center crop
)
print("Input blob shape: {}\n".format(input_blob.shape))

Please, pay attention at the preprocessing order in the cv2.dnn.blobFromImage function. Firstly, the mean value is subtracted and only then pixel values are multiplied by the defined scale. Therefore, to reproduce the image preprocessing pipeline from the TF mobilenet.preprocess_input function, we multiply mean by 127.5.

As a result, 4-dimensional input_blob was obtained:

Input blob shape: (1, 3, 224, 224)

# set OpenCV DNN input
opencv_net.setInput(preproc_img)
# OpenCV DNN inference
out = opencv_net.forward()
print("OpenCV DNN prediction: \n")
print("* shape: ", out.shape)
# get the predicted class ID
imagenet_class_id = np.argmax(out)
# get confidence
confidence = out[0][imagenet_class_id]
print("* class ID: {}, label: {}".format(imagenet_class_id, imagenet_labels[imagenet_class_id]))
print("* confidence: {:.4f}\n".format(confidence))

After the above code execution we will get the following output:

OpenCV DNN prediction:
* shape: (1, 1000)
* class ID: 335, label: fox squirrel, eastern fox squirrel, Sciurus niger
* confidence: 0.9525
# inference
preproc_img = preproc_img.transpose(0, 2, 3, 1)
print("TF input blob shape: {}\n".format(preproc_img.shape))
out = original_net(preproc_img)
print("\nTensorFlow model prediction: \n")
print("* shape: ", out.shape)
# get the predicted class ID
imagenet_class_id = np.argmax(out)
print("* class ID: {}, label: {}".format(imagenet_class_id, imagenet_labels[imagenet_class_id]))
# get confidence
confidence = out[0][imagenet_class_id]
print("* confidence: {:.4f}".format(confidence))

To fit TF model input, input_blob was transposed:

TF input blob shape: (1, 224, 224, 3)

TF inference results are the following:

TensorFlow model prediction:
* shape: (1, 1000)
* class ID: 335, label: fox squirrel, eastern fox squirrel, Sciurus niger
* confidence: 0.9525

As it can be seen from the experiments OpenCV and TF inference results are equal.

Evaluation of the Models

The proposed in dnn/samples dnn_model_runner module allows to run the full evaluation pipeline on the ImageNet dataset and test execution for the following TensorFlow classification models:

This list can be also extended with further appropriate evaluation pipeline configuration.

Evaluation Mode

To below line represents running of the module in the evaluation mode:

python -m dnn_model_runner.dnn_conversion.tf.classification.py_to_py_cls --model_name <tf_cls_model_name>

Chosen from the list classification model will be read into OpenCV cv.dnn_Net object. Evaluation results of TF and OpenCV models (accuracy, inference time, L1) will be written into the log file. Inference time values will be also depicted in a chart to generalize the obtained model information.

Necessary evaluation configurations are defined in the test_config.py and can be modified in accordance with actual paths of data location::

@dataclass
class TestClsConfig:
batch_size: int = 50
frame_size: int = 224
img_root_dir: str = "./ILSVRC2012_img_val"
# location of image-class matching
img_cls_file: str = "./val.txt"
bgr_to_rgb: bool = True

The values from TestClsConfig can be customized in accordance with chosen model.

To initiate the evaluation of the TensorFlow MobileNet, run the following line:

python -m dnn_model_runner.dnn_conversion.tf.classification.py_to_py_cls --model_name mobilenet

After script launch, the log file with evaluation data will be generated in dnn_model_runner/dnn_conversion/logs:

===== Running evaluation of the model with the following params:
* val data location: ./ILSVRC2012_img_val
* log file location: dnn_model_runner/dnn_conversion/logs/TF_mobilenet_log.txt

Test Mode

The below line represents running of the module in the test mode, namely it provides the steps for the model inference:

python -m dnn_model_runner.dnn_conversion.tf.classification.py_to_py_cls --model_name <tf_cls_model_name> --test True --default_img_preprocess <True/False> --evaluate False

Here default_img_preprocess key defines whether you'd like to parametrize the model test process with some particular values or use the default values, for example, scale, mean or std.

Test configuration is represented in test_config.py TestClsModuleConfig class:

@dataclass
class TestClsModuleConfig:
cls_test_data_dir: str = "../data"
test_module_name: str = "classification"
test_module_path: str = "classification.py"
input_img: str = os.path.join(cls_test_data_dir, "squirrel_cls.jpg")
model: str = ""
frame_height: str = str(TestClsConfig.frame_size)
frame_width: str = str(TestClsConfig.frame_size)
scale: str = "1.0"
mean: List[str] = field(default_factory=lambda: ["0.0", "0.0", "0.0"])
std: List[str] = field(default_factory=list)
crop: str = "False"
rgb: str = "True"
rsz_height: str = ""
rsz_width: str = ""
classes: str = os.path.join(cls_test_data_dir, "dnn", "classification_classes_ILSVRC2012.txt")

The default image preprocessing options are defined in default_preprocess_config.py. For instance, for MobileNet:

tf_input_blob = {
"mean": ["127.5", "127.5", "127.5"],
"scale": str(1 / 127.5),
"std": [],
"crop": "True",
"rgb": "True"
}

The basis of the model testing is represented in samples/dnn/classification.py. classification.py can be executed autonomously with provided converted model in --input and populated parameters for cv.dnn.blobFromImage.

To reproduce from scratch the described in "Model Conversion Pipeline" OpenCV steps with dnn_model_runner execute the below line:

python -m dnn_model_runner.dnn_conversion.tf.classification.py_to_py_cls --model_name mobilenet --test True --default_img_preprocess True --evaluate False

The network prediction is depicted in the top left corner of the output window:

tf_mobilenet_opencv_test_res.jpg
TF MobileNet OpenCV inference output