/WSS_POLE

This is the official implementation of our PrOmpt cLass lEarning (POLE).

Primary LanguagePythonMIT LicenseMIT

WSS_POLE

This is a sample visualization from our PrOmpt cLass lEarning (POLE). Full training and inference codes will be released upon paper acceptance.

Setup

  1. Clone Repository:
git clone https://github.com/Ruxie189/WSS_POLE.git
  1. Create a new python virtual environment as:
python -m venv path/to/env/
  1. Activate environment:
source path/to/env/bin/activate
  1. Install dependencies:
pip install -r reqs.txt

Running the Notebook

Open a terminal in the same directory as that of the cloned repository and run:

jupyter-notebook

Open sim_test.ipynb to find the codes for visualizing CLIP similarity. Running the notebook is mostly self explanatory and further instructions have been provided inside the notebook.

Citations

@inproceedings{murugesan2024prompting,
  title={Prompting classes: Exploring the Power of Prompt Class Learning in Weakly Supervised Semantic Segmentation},
  author={Murugesan, Balamurali and Hussain, Rukhshanda and Bhattacharya, Rajarshi and Ben Ayed, Ismail and Dolz, Jose},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={291--302},
  year={2024}
}