STAM - Pytorch
Implementation of STAM (Space Time Attention Model), yet another pure and simple SOTA attention model that bests all previous models in video classification. This corroborates the finding of TimeSformer. Attention is all we need.
Install
$ pip install stam-pytorch
Usage
import torch
from stam_pytorch import STAM
model = STAM(
dim = 512,
image_size = 256, # size of image
patch_size = 32, # patch size
num_frames = 5, # number of image frames, selected out of video
space_depth = 12, # depth of vision transformer
space_heads = 8, # heads of vision transformer
space_mlp_dim = 2048, # feedforward hidden dimension of vision transformer
time_depth = 6, # depth of time transformer (in paper, it was shallower, 6)
time_heads = 8, # heads of time transformer
time_mlp_dim = 2048, # feedforward hidden dimension of time transformer
num_classes = 100, # number of output classes
space_dim_head = 64, # space transformer head dimension
time_dim_head = 64, # time transformer head dimension
dropout = 0., # dropout
emb_dropout = 0. # embedding dropout
)
frames = torch.randn(2, 5, 3, 256, 256) # (batch x frames x channels x height x width)
pred = model(frames) # (2, 100)
Citations
@misc{sharir2021image,
title = {An Image is Worth 16x16 Words, What is a Video Worth?},
author = {Gilad Sharir and Asaf Noy and Lihi Zelnik-Manor},
year = {2021},
eprint = {2103.13915},
archivePrefix = {arXiv},
primaryClass = {cs.CV}
}
191 Dec 31, 2022
2 Jun 5, 2022
58 Dec 21, 2022
159 Dec 22, 2022
54 Nov 21, 2022
8 Oct 29, 2022
272 Dec 23, 2022
1.4k Jan 1, 2023
602 Jan 3, 2023
1k Dec 31, 2022
85 Dec 30, 2022
3 Jan 7, 2022
1 Mar 29, 2022
98 Nov 16, 2022
95 Oct 24, 2022
456 Dec 12, 2022