yas-sim / openvino-segment-anything-interactive-demo

Segment Anything Model (SAM) interactive demo with OpenVINO

Segment Anything Model (SAM) interactive demo with OpenVINO

Description

This demo takes an image file and runs a SAM model using OpenVINO.
You can point to the input image with the mouse, and the demo program shows the segmentation result on the image.
This demo supports following models from yformer/EfficientSAM GitHub repo:

• efficient-sam-vitt
• efficient-sam-vits

Programs / Files

#	File name	Description
1	download_and_convert.py	Clone the EfficientSAM GitHub repo and convert the model into OpenVINO IR model.
2	efficientsam.py	Run the EfficientSAM interactive demo with OpenVINO

How to run

1. Install prerequisites

python -m venv venv
venv/Scripts/activate
python -m pip install -U pip
pip install -U setuptools wheel
pip install -r requirements.txt

2. Download the model

python download_and_convert.py

3. Run the demo Point and click on the image to kick inferencing.

python efficientsam.py
  or
python efficientsam.py -i <image_file_name>

Command line options:

options:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        Path to an input image file
  -n NUM_POINTS, --num_points NUM_POINTS
                        Number of points for an inference (default:1)
  -f, --full_screen     Full screen mode

Demo

python .\efficientsam.py -i .\trolley.jpg -n 3 -f

Test environment

• Windows 11
• Python 3.10.3
• OpenVINO 2023.3.0