Image In-painting with OpenVINO™

This tutorial is also available as a Jupyter notebook that can be cloned directly from GitHub. See the installation guide for instructions to run this tutorial locally on Windows, Linux or macOS. To run without installing anything, click the launch binder button.

Binder Github

This notebook demonstrates how to use an image in-painting model with OpenVINO, using GMCNN model from Open Model Zoo. This model, given a tampered image, is able to create something very similar to the original image. The Following pipeline will be used in this notebook. pipeline

import sys
from pathlib import Path

import cv2
import matplotlib.pyplot as plt
import numpy as np
from openvino.runtime import Core

import notebook_utils as utils

Download the Model

Models can be downloaded with omz_downloader - a command-line tool for downloading models from the Open Model Zoo. gmcnn-places2-tf is the omz name for the considered model. You can find the names of all available models here and here. The selected model comes from the public directory, which means it must be converted into Intermediate Representation (IR). This step is skipped if the model is already downloaded.

# A directory where the model will be downloaded.
base_model_dir = "model"
# The name of the model from Open Model Zoo.
model_name = "gmcnn-places2-tf"

model_path = Path(f"{base_model_dir}/public/{model_name}/{model_name}/frozen_model.pb")
if not model_path.exists():
    download_command = f"omz_downloader " \
                       f"--name {model_name} " \
                       f"--output_dir {base_model_dir}"
    ! $download_command
    print("Already downloaded")
################|| Downloading gmcnn-places2-tf ||################

========== Downloading model/public/gmcnn-places2-tf/

========== Unpacking model/public/gmcnn-places2-tf/

Convert Tensorflow model to OpenVINO IR format

The pre-trained model is in TensorFlow format. To use it with OpenVINO, convert it to OpenVINO IR format. To do this, use Model Converter (omz_converter), which is another command-line tool. Then, specify the precision for FP32 but it can be FP16 as well. This step is also skipped if the model is already converted.

precision = "FP32"
ir_path = Path(f"{base_model_dir}/public/{model_name}/{precision}/{model_name}.xml")

# Run Model Optimizer if the OpenVINO IR model file does not exist.
if not ir_path.exists():
    print("Exporting TensorFlow model to IR... This may take a few minutes.")
    convert_command = f"omz_converter " \
                      f"--name {model_name} " \
                      f"--download_dir {base_model_dir} " \
                      f"--precisions {precision}"
    ! $convert_command
    print("IR model already exists.")
Exporting TensorFlow model to IR... This may take a few minutes.
========== Converting gmcnn-places2-tf to IR (FP32)
Conversion command: /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/.venv/bin/python -- /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/.venv/bin/mo --framework=tf --data_type=FP32 --output_dir=model/public/gmcnn-places2-tf/FP32 --model_name=gmcnn-places2-tf --input=Placeholder,Placeholder_1 --input_model=model/public/gmcnn-places2-tf/gmcnn-places2-tf/frozen_model.pb --output=Minimum '--layout=Placeholder(NHWC),Placeholder_1(NHWC)' '--input_shape=[1, 512, 680, 3],[1, 512, 680, 1]'

Model Optimizer arguments:
Common parameters:
    - Path to the Input Model:  /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/notebooks/215-image-inpainting/model/public/gmcnn-places2-tf/gmcnn-places2-tf/frozen_model.pb
    - Path for generated IR:    /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/notebooks/215-image-inpainting/model/public/gmcnn-places2-tf/FP32
    - IR output name:   gmcnn-places2-tf
    - Log level:    ERROR
    - Batch:    Not specified, inherited from the model
    - Input layers:     Placeholder,Placeholder_1
    - Output layers:    Minimum
    - Input shapes:     [1, 512, 680, 3],[1, 512, 680, 1]
    - Source layout:    Not specified
    - Target layout:    Not specified
    - Layout:   Placeholder(NHWC),Placeholder_1(NHWC)
    - Mean values:  Not specified
    - Scale values:     Not specified
    - Scale factor:     Not specified
    - Precision of IR:  FP32
    - Enable fusing:    True
    - User transformations:     Not specified
    - Reverse input channels:   False
    - Enable IR generation for fixed input shape:   False
    - Use the transformations config file:  None
Advanced parameters:
    - Force the usage of legacy Frontend of Model Optimizer for model conversion into IR:   False
    - Force the usage of new Frontend of Model Optimizer for model conversion into IR:  False
TensorFlow specific parameters:
    - Input model in text protobuf format:  False
    - Path to model dump for TensorBoard:   None
    - List of shared libraries with TensorFlow custom layers implementation:    None
    - Update the configuration file with input/output node names:   None
    - Use configuration file used to generate the model with Object Detection API:  None
    - Use the config file:  None
OpenVINO runtime found in:  /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/.venv/lib/python3.8/site-packages/openvino
OpenVINO runtime version:   2022.2.0-7713-af16ea1d79a-releases/2022/2
Model Optimizer version:    2022.2.0-7713-af16ea1d79a-releases/2022/2
[ SUCCESS ] Generated IR version 11 model.
[ SUCCESS ] XML file: /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/notebooks/215-image-inpainting/model/public/gmcnn-places2-tf/FP32/gmcnn-places2-tf.xml
[ SUCCESS ] BIN file: /opt/home/k8sworker/cibuilds/ov-notebook/OVNotebookOps-275/.workspace/scm/ov-notebook/notebooks/215-image-inpainting/model/public/gmcnn-places2-tf/FP32/gmcnn-places2-tf.bin
[ SUCCESS ] Total execution time: 10.55 seconds.
[ SUCCESS ] Memory consumed: 612 MB.
[ INFO ] The model was converted to IR v11, the latest model format that corresponds to the source DL framework input/output format. While IR v11 is backwards compatible with OpenVINO Inference Engine API v1.0, please use API v2.0 (as of 2022.1) to take advantage of the latest improvements in IR v11.
Find more information about API v2.0 and IR v11 at

Load the model

Now, load the OpenVINO IR model and perform as follows:

  1. Initialize OpenVINO Runtime (Core).

  2. Read the network from *.bin and *.xml files (weights and architecture)

  3. Compile the model for the “CPU”.

  4. Get input and output nodes.

Only a few lines of code are required to run the model:

core = Core()

# Read the model.xml and weights file
model = core.read_model(model=ir_path)
# Load the model on to the CPU
compiled_model = core.compile_model(model=model, device_name="CPU")
# Store the input and output nodes
input_layer = compiled_model.input(0)
output_layer = compiled_model.output(0)

Determine the input shapes of the model

Note that both input shapes are the same. However, the second input has 1 channel (monotone).

N, H, W, C = input_layer.shape

Create a square mask

Next, create a single channeled mask that will be laid on top of the original image.

def create_mask(image_width, image_height, size_x=30, size_y=30, number=1):
    Create a square mask of defined size on a random location.

    :param: image_width: width of the image
    :param: image_height: height of the image
    :param: size: size in pixels of one side
            mask: grayscale float32 mask of size shaped [image_height, image_width, 1]

    mask = np.zeros((image_height, image_width, 1), dtype=np.float32)
    for _ in range(number):
        start_x = np.random.randint(image_width - size_x)
        start_y = np.random.randint(image_height - size_y)
                      pt1=(start_x, start_y),
                      pt2=(start_x + size_x, start_y + size_y),
                      color=(1, 1, 1),
    return mask
# Generate a square mask of size WxH with number of "holes".
mask = create_mask(image_width=W, image_height=H, size_x=50, size_y=50, number=15)
# This mask will be laid over the input image as noise.
plt.figure(figsize=(16, 12))
plt.imshow(cv2.cvtColor(mask, cv2.COLOR_BGR2RGB));

Load and Resize the Image

This image will be altered by using the mask. You can process any image you like. Just change the URL below.

# Download an image.
url = ""
image_file = utils.download_file(
    url, filename="laptop.png", directory="data", show_progress=False, silent=True, timeout=30
assert Path(image_file).exists()

# Read the image.
image = cv2.imread("data/laptop.png")
# Resize the image to meet network expected input sizes.
resized_image = cv2.resize(src=image, dsize=(W, H), interpolation=cv2.INTER_AREA)
plt.figure(figsize=(16, 12))
plt.imshow(cv2.cvtColor(resized_image, cv2.COLOR_BGR2RGB));

Generating the Masked Image

This multiplication of the image and the mask gives the result of the masked image layered on top of the original image. The masked_image will be the first input to the GMCNN model.

# Generating a masked image.
masked_image = (resized_image * (1 - mask) + 255 * mask).astype(np.uint8)
plt.figure(figsize=(16, 12))
plt.imshow(cv2.cvtColor(masked_image, cv2.COLOR_BGR2RGB));


The model expects the input dimensions to be NHWC.

  • masked_image.shape = (512,680,3) —–> model expects = (1,512,680,3)

  • resized_mask.shape = (512,680,1) —–> model expects = (1,512,680,1)

masked_image = masked_image[None, ...]
mask = mask[None, ...]


Do inference with the given masked image and the mask. Then, show the restored image.

result = compiled_model([masked_image, mask])[output_layer]
result = result.squeeze().astype(np.uint8)
plt.figure(figsize=(16, 12))
plt.imshow(cv2.cvtColor(result, cv2.COLOR_BGR2RGB))
<matplotlib.image.AxesImage at 0x7f94c83cd6a0>

Save the Restored Image

Save the restored image to the data directory to download it.

cv2.imwrite("data/laptop_restored.png", result)