sa

slot-attention

Implementation of Slot Attention from GoogleAI

Showing:

Popularity

Downloads/wk

0

GitHub Stars

209

Maintenance

Last Commit

8mos ago

Contributors

3

Package

Dependencies

1

License

MIT

Categories

Readme

Slot Attention

Implementation of Slot Attention from the paper 'Object-Centric Learning with Slot Attention' in Pytorch. Here is a video that describes what this network can do.

Update: The official repository has been released here

Install

$ pip install slot_attention

Usage

import torch
from slot_attention import SlotAttention

slot_attn = SlotAttention(
    num_slots = 5,
    dim = 512,
    iters = 3   # iterations of attention, defaults to 3
)

inputs = torch.randn(2, 1024, 512)
slot_attn(inputs) # (2, 5, 512)

After training, the network is reported to be able to generalize to slightly different number of slots (clusters). You can override the number of slots used by the num_slots keyword in forward.

slot_attn(inputs, num_slots = 8) # (2, 8, 512)

Citation

@misc{locatello2020objectcentric,
    title = {Object-Centric Learning with Slot Attention},
    author = {Francesco Locatello and Dirk Weissenborn and Thomas Unterthiner and Aravindh Mahendran and Georg Heigold and Jakob Uszkoreit and Alexey Dosovitskiy and Thomas Kipf},
    year = {2020},
    eprint = {2006.15055},
    archivePrefix = {arXiv},
    primaryClass = {cs.LG}
}

Rate & Review

Great Documentation0
Easy to Use0
Performant0
Highly Customizable0
Bleeding Edge0
Responsive Maintainers0
Poor Documentation0
Hard to Use0
Slow0
Buggy0
Abandoned0
Unwelcoming Community0
100