Blur Image via Prompts

Introduction

We implement a module that blurs objects (in an image) determined by the user (text) prompts. While constructing the module, we utilized the pretrained models OWLViT-v2 and mobile-SAM provided by HuggingFace and ultralytics, respectively. The demo is accessible at the HuggingFace space.

Setting Up the Environment

1. Install Conda, if not already installed.
2. Clone the repository

   git clone https://github.com/byrkbrk/blurring-image-via-prompts.git
   

3. Change the directory:

   cd blurring-image-via-prompts
   

4. For macos, run:

   conda env create -f blurring-via-prompts_macos.yaml
   

   For linux or windows, run:

   conda env create -f blurring-via-prompts_linux.yaml
   

5. Activate the environment:

   conda activate blurring-via-prompts
   

Blurring Objects

Check it out how to use:

python3 blur.py -h

Output:

Blurs image based on given text prompts

positional arguments:
  image_name            Name of the image file that be processed. Image file
                        must be in `images-to-blur` folder
  text_prompts          Text prompts for the objects that get blurred

options:
  -h, --help            show this help message and exit
  --blur_intensity BLUR_INTENSITY
                        Intensity of the blur that be applied. Default: 50
  --image_size IMAGE_SIZE [IMAGE_SIZE ...]
                        Size (width, height) to which the image be
                        transformed. Default: None
  --device DEVICE       Device that be used during inference. Default: None

Example usage

python3 blur.py dogs.jpg "jacket"

The output image (see below, on the right) will be saved into blurred-images folder.

Blurring via Gradio

To run the gradio app on your local computer, execute

python3 app.py

Then, visit the url http://127.0.0.1:7860 to open the interface seen below.

Example usage

See the display below for an example usage of the module via Gradio for the image hat_sunglasses.jpg (foundable in the directory images-to-blur).