Unsupervised Object Discovery

GitHub license Github commits made-with-python GitHub issues GitHub pull-requests

This is an implementation of several unsupervised object discovery models (Slot Attention, SLATE, GNM) in PyTorch with pre-trained models.

An example of the results of the Slot Attention model on the CLEVR dataset.

Unsupervised object discovery is the task of segmenting objects in an image without ever being given a label indicating where the objects are. You can view more visualization like the one on this page in the visualizations section of the README.

Project link: https://github.com/HHousen/object-discovery-pytorch