Computer Vision Projects

computervision #panopticsegmentation #instancesegmentation

Projects

1. Panoptic Segmentation
   • Notebook: panoptic_segmentation.ipynb (Notebook link
   • Description: Identifies and extracts specific objects from images using panoptic segmentation, then composites those objects onto a transparent background.
   • Platform - detectron2 by Facebook Research
   • Example:
     • Original Image: /images/originals/example1.jpg (Image Source: COCO Dataset) )
     • Inferred Image: /images/output/example_panoptic.jpg
     • Masked Image: /images/outputs/example_masked.jpg

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/cv-object-segmentation.git
   cd cv-object-segmentation
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

Usage

1. Place your input images in the root/images/originals/ directory.
2. Run the relevant Jupyter notebook.
3. Adjust the class_names variable to specify the objects you want to extract.
4. View the results in the root/images/masked/ directory.

License

This project is licensed under the MIT License - see the LICENSE file for details.