modnet-photographic-portrait-matting

Use Case and High-Level Description

The modnet-photographic-portrait-matting model is a lightweight matting objective decomposition network (MODNet) for photographic portrait matting in real-time with a single input image with MobileNetV2 backbone. The model is pre-trained in PyTorch* framework and converted to ONNX* format.

More details provided in the paper and repository.

Specification

Metric

Value

Type

Background Matting

GFlops

31.1564

MParams

6.4597

Source framework

PyTorch*

Accuracy

Accuracy measured on the HumanMatting dataset

Metric

Mean value

Std value

MAD

5.21

5.13

MSE

727.95

1196.28

  • MAD - mean of absolute difference

  • MSE - mean squared error.

Input

Original Model

Image, name: input, shape: 1, 3, 512, 512, format: B, C, H, W, where:

  • B - batch size

  • C - number of channels

  • H - image height

  • W - image width

Expected color order: RGB. Mean values - [127.5, 127.5, 127.5], scale value - 127.5.

Converted Model

Image, name: input, shape: 1, 3, 512, 512, format: B, C, H, W, where:

  • B - batch size

  • C - number of channels

  • H - image height

  • W - image width

Expected color order: BGR.

Output

Original model

Alpha matte with values in [0, 1] range. Name: output Shape: 1, 1, 512, 512, format: B, C, H, W, where:

  • B - batch size

  • C - number of channels

  • H - image height

  • W - image width

Converted model

Alpha matte with values in [0, 1] range. Name: output Shape: 1, 1, 512, 512, format: B, C, H, W, where:

  • B - batch size

  • C - number of channels

  • H - image height

  • W - image width

Download a Model and Convert it into OpenVINO™ IR Format

You can download models and if necessary convert them into OpenVINO™ IR format using the Model Downloader and other automation tools as shown in the examples below.

An example of using the Model Downloader:

omz_downloader --name <model_name>

An example of using the Model Converter:

omz_converter --name <model_name>

Demo usage

The model can be used in the following demos provided by the Open Model Zoo to show its capabilities: