mhamilton723/STEGO
An unsupervised semantic segmentation method that distills feature correspondences into discrete segmentation masks using deep learning.
Not currently ranked — collecting fresh signals.
star history
STEGO is a deep learning research project implementing an unsupervised semantic segmentation approach. The method learns to segment images into meaningful categories without requiring labeled training data by distilling feature correspondences from a pre-trained vision transformer. The implementation uses PyTorch and includes training, evaluation pipelines, and a Colab demo for reproducing the ICLR 2022 paper results.
Frequently asked
- What is mhamilton723/STEGO?
- An unsupervised semantic segmentation method that distills feature correspondences into discrete segmentation masks using deep learning.
- Is STEGO open source?
- Yes — mhamilton723/STEGO is open source, released under the MIT license.
- What language is STEGO written in?
- mhamilton723/STEGO is primarily written in Jupyter Notebook.
- How popular is STEGO?
- mhamilton723/STEGO has 791 stars on GitHub.
- Where can I find STEGO?
- mhamilton723/STEGO is on GitHub at https://github.com/mhamilton723/STEGO.