doi
stringlengths 9
10
| chunk-id
stringlengths 1
4
| chunk
stringlengths 1
1.57k
| id
stringlengths 9
10
| title
stringlengths 10
127
| summary
stringlengths 581
1.92k
| source
stringlengths 30
31
| authors
list | categories
list | comment
stringclasses 28
values | journal_ref
stringclasses 1
value | primary_category
stringclasses 6
values | published
stringlengths 8
8
| updated
stringlengths 8
8
| references
list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2102.05095
|
171
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
172
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
173
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
174
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
175
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
176
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit>z(`)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
177
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
178
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
179
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
180
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
181
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
182
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
183
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
184
|
sha1_base64="vr9M27hwkmJn9HyewcyBFvyrmCg=">AAAB+XicbVBNS8NAEN3Ur1q/oh69LBahXkoigh6LXjxWsB/QxLLZTtqlm03Y3RRqyD/x4kERr/4Tb/4bt20O2vpg4PHeDDPzgoQzpR3n2yqtrW9sbpW3Kzu7e/sH9uFRW8WppNCiMY9lNyAKOBPQ0kxz6CYSSBRw6ATj25nfmYBULBYPepqAH5GhYCGjRBupb9uZF4T4KX/Mah5wfp737apTd+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRml
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
185
|
+bAq8QtSBUVaPbtL28Q0zQCoSknSvVcJ9F+RqRmlENe8VIFCaFjMoSeoYJEoPxsfnmOz4wywGEsTQmN5+rviYxESk2jwHRGRI/UsjcT//N6qQ6v/YyJJNUg6GJRmHKsYzyLAQ+YBKr51BBCJTO3YjoiklBtwqqYENzll1dJ+6LuOnX3/rLauCniKKMTdIpqyEVXqIHuUBO1EEUT9Ixe0ZuVWS/Wu/WxaC1Zxcwx+gPr8wcJUZNB</latexit>z(` 1)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
186
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
187
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
188
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
189
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
190
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
191
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
192
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
193
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
194
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit>z(` 1)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
195
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
196
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
197
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
198
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
199
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
200
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
201
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
202
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
203
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit>z(` 1)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
204
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
205
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
206
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
207
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
208
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
209
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
210
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
211
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
212
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit>z(` 1)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
213
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
214
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
215
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
216
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
217
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
218
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
219
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
220
|
sha1_base64="PqFOkq34IvzyoSmk0/rjIYa/Lb0=">AAAB+3icbVBNS8NAEN3Ur1q/Yj16WSxCPVgSEfRY9OKxgv2ANpbNdtIu3WzC7kasIX/FiwdFvPpHvPlv3LY5aOuDgcd7M8zM82POlHacb6uwsrq2vlHcLG1t7+zu2fvllooSSaFJIx7Jjk8UcCagqZnm0IklkNDn0PbH11O//QBSsUjc6UkMXkiGggWMEm2kvl1Oe36An7L7tNoDzk/dk6xvV5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lR
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
221
|
5yaMwNeJm5OKihHo29/9QYRTUIQmnKiVNd1Yu2lRGpGOWSlXqIgJnRMhtA1VJAQlJfObs/wsVEGOIikKaHxTP09kZJQqUnom86Q6JFa9Kbif1430cGllzIRJxoEnS8KEo51hKdB4AGTQDWfGEKoZOZWTEdEEqpNXCUTgrv48jJpndVcp+benlfqV3kcRXSIjlAVuegC1dENaqAmougRPaNX9GZl1ov1bn3MWwtWPnOA/sD6/AHtlpOz</latexit>
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
222
|
Figure 1. The video self-attention blocks that we investigate in this work. Each attention layer implements self-attention (Vaswani et al.,
2017b) on a specified spatiotemporal neighborhood of frame-level patches (see Figure 2 for a visualization of the neighborhoods). We use
residual connections to aggregate information from different attention layers within each block. A 1-hidden-layer MLP is applied at the
end of each block. The final model is constructed by repeatedly stacking these blocks on top of each other.
p= 1;:::;N denoting spatial locations and t= 1;:::;F
depicting an index over frames.
Linear embedding. We linearly map each patch x(p;t)into
an embedding vector z(0)
(p;t)2RDby means of a learnable
matrixE2RD3P2:
z(0)
(p;t)=Ex(p;t)+epos
(p;t)(1)
where epos
(p;t)2RDrepresents a learnable positional embedding added to encode the spatiotemporal position of each
patch. The resulting sequence of embedding vectors z(0)
(p;t)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
223
|
patch. The resulting sequence of embedding vectors z(0)
(p;t)
forp= 1;:::;N , andt= 1;:::;F represents the input to
the Transformer, and plays a role similar to the sequences of
embedded words that are fed to text Transformers in NLP.
As in the original BERT Transformer (Devlin et al., 2018),
we add in the first position of the sequence a special learnable vector z(0)
(0;0)2RDrepresenting the embedding of the
classification token.
Query-Key-Value computation. Our Transformer consists
ofLencoding blocks. At each block `, a query/key/value
vector is computed for each patch from the representation
z(` 1)
(p;t)encoded by the preceding block:
q(`;a)
(p;t)=W(`;a)
QLN
z(` 1)
(p;t)
2RDh(2)
k(`;a)
(p;t)=W(`;a)
KLN
z(` 1)
(p;t)
2RDh(3)
v(`;a)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
224
|
KLN
z(` 1)
(p;t)
2RDh(3)
v(`;a)
(p;t)=W(`;a)
VLN
z(` 1)
(p;t)
2RDh(4)where LN() denotes LayerNorm (Ba et al., 2016), a=
1;:::;Ais an index over multiple attention heads and A
denotes the total number of attention heads. The latent
dimensionality for each attention head is set to Dh=D=A.
Self-attention computation. Self-attention weights are
computed via dot-product. The self-attention weights
(`;a)
(p;t)2RNF+1for query patch (p;t)are given by:
(`;a)
(p;t)= SM0
@q(`;a)
(p;t)pDh>
"
k(`;a)
(0;0)n
k(`;a)
(p0;t0)o
p0=1;:::;N
t0=1;:::;F#1
A
(5)
where SMdenotes the softmax activation function. Note
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
225
|
p0=1;:::;N
t0=1;:::;F#1
A
(5)
where SMdenotes the softmax activation function. Note
that when attention is computed over one dimension only
(e.g., spatial-only or temporal-only), the computation is
significantly reduced. For example, in the case of spatial
attention, only N+ 1 query-key comparisons are made,
using exclusively keys from the same frame as the query:
(`;a)space
(p;t)= SM0
@q(`;a)
(p;t)pDh>
k(`;a)
(0;0)n
k(`;a)
(p0;t)o
p0=1;:::;N1
A:
(6)
Encoding. The encoding z(`)
(p;t)at block`is obtained by
first computing the weighted sum of value vectors using
self-attention coefficients from each attention head:
Is Space-Time Attention All You Need for Video Understanding?
frame tframe t -
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
226
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
227
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
228
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
229
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
230
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
231
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
232
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
233
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
234
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit>frame
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
235
|
t +
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
236
|
<latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
237
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
238
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
239
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
240
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
241
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
242
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit><latexit
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
243
|
sha1_base64="AU/vq+DAgMm1x9RhUc20aA7BJwY=">AAAB7XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0GPRi8cK9gPaUDabTbt2kw27E6GE/gcvHhTx6v/x5r9x2+agrQ8GHu/NMDMvSKUw6LrfTmltfWNzq7xd2dnd2z+oHh61jco04y2mpNLdgBouRcJbKFDybqo5jQPJO8H4duZ3nrg2QiUPOEm5H9NhIiLBKFqp3Q+5RDqo1ty6OwdZJV5BalCgOah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
244
|
ah+9UPFspgnyCQ1pue5Kfo51SiY5NNKPzM8pWxMh7xnaUJjbvx8fu2UnFklJJHSthIkc/X3RE5jYyZxYDtjiiOz7M3E/7xehtG1n4skzZAnbLEoyiRBRWavk1BozlBOLKFMC3srYSOqKUMbUMWG4C2/vEraF3XPrXv3l7XGTRFHGU7gFM7BgytowB00oQUMHuEZXuHNUc6L8+58LFpLTjFzDH/gfP4AkYyPHA==</latexit>Space
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
245
|
Attention (S) Joint Space-Time Attention (ST)Divided Space-Time Attention (T+S) Sparse Local Global Attention (L+G) Axial Attention (T+W+H)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
246
|
Figure 2. Visualization of the five space-time self-attention schemes studied in this work. Each video clip is viewed as a sequence of
frame-level patches with a size of 1616pixels. For illustration, we denote in blue the query patch and show in non-blue colors its
self-attention space-time neighborhood under each scheme. Patches without color are not used for the self-attention computation of the
blue patch. Multiple colors within a scheme denote attentions separately applied along different dimensions (e.g., space and time for
(T+S)) or over different neighborhoods (e.g., for (L+G)). Note that self-attention is computed for every single patch in the video clip, i.e.,
every patch serves as a query. We also note that although the attention pattern is shown for only two adjacent frames, it extends in the
same fashion to all frames of the clip.
s(`;a)
(p;t)= (`;a)
(p;t);(0;0)v(`;a)
(0;0)+NX
p0=1FX
t0=1 (`;a)
(p;t);(p0;t0)v(`;a)
(p0;t0):
(7)
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
247
|
(p;t);(p0;t0)v(`;a)
(p0;t0):
(7)
Then, the concatenation of these vectors from all heads
is projected and passed through an MLP, using residual
connections after each operation:
z0(`)
(p;t)=WO2
664s(`;1)
(p;t)
...
s(`;A)
(p;t)3
775+z(` 1)
(p;t)(8)
z(`)
(p;t)= MLP
LN
z0(`)
(p;t)
+z0(`)
(p;t): (9)
Classification embedding. The final clip embedding is
obtained from the final block for the classification token:
y= LN
z(L)
(0;0)
2RD: (10)
On top of this representation we append a 1-hidden-layer
MLP, which is used to predict the final video classes.
Space-Time Self-Attention Models. We can reduce the
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
248
|
MLP, which is used to predict the final video classes.
Space-Time Self-Attention Models. We can reduce the
computational cost by replacing the spatiotemporal attention of Eq. 5 with spatial attention within each frame only
(Eq. 6). However, such a model neglects to capture temporaldependencies across frames. As shown in our experiments,
this approach leads to degraded classification accuracy compared to full spatiotemporal attention, especially on benchmarks where strong temporal modeling is necessary.
We propose a more efficient architecture for spatiotemporal
attention, named “Divided Space-Time Attention” (denoted
with T+S), where temporal attention and spatial attention
are separately applied one after the other. This architecture
is compared to that of Space and Joint Space-Time attention
in Fig. 1. A visualization of the different attention models
on a video example is given in Fig. 2. For Divided Attention,
within each block `, we first compute temporal attention by
comparing each patch (p;t)with all the patches at the same
spatial location in the other frames:
(`;a)time
(p;t)= SM0
@q(`;a)
(p;t)pDh>
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
249
|
(`;a)time
(p;t)= SM0
@q(`;a)
(p;t)pDh>
k(`;a)
(0;0)n
k(`;a)
(p;t0)o
t0=1;:::;F1
A:
(11)
The encoding z0(`)time
(p;t)resulting from the application of
Eq. 8 using temporal attention is then fed back for spatial
attention computation instead of being passed to the MLP. In
other words, new key/query/value vectors are obtained from
z0(`)time
(p;t)and spatial attention is then computed using Eq. 6.
Finally, the resulting vector z0(`)space
(p;t)is passed to the MLP
of Eq. 9 to compute the final encoding z(`)
(p;t)of the patch at
block`. For the model of divided attention, we learn distinct query/key/value matrices fW(`;a)
Qtime;W(`;a)
Ktime;W(`;a)
Vtimeg
Is Space-Time Attention All You Need for Video Understanding?
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
250
|
Qtime;W(`;a)
Ktime;W(`;a)
Vtimeg
Is Space-Time Attention All You Need for Video Understanding?
Attention Params K400 SSv2
Space 85.9M 76.9 36.6
Joint Space-Time 85.9M 77.4 58.5
Divided Space-Time 121.4M 78.0 59.5
Sparse Local Global 121.4M 75.9 56.3
Axial 156.8M 73.5 56.2
Table 1. Video-level accuracy for different space-time attention
schemes in TimeSformer. We evaluate the models on the validation sets of Kinetics-400 (K400), and Something-Something-V2
(SSv2). We observe that divided space-time attention achieves the
best results on both datasets.
andfW(`;a)
Qspace;W(`;a)
Kspace;W(`;a)
Vspacegover the time and space
dimensions. Note that compared to the (NF+ 1) comparisons per patch needed by the joint spatiotemporal attention
model of Eq. 5, Divided Attention performs only (N+F+2)
comparisons per patch. Our experiments demonstrate that
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
251
|
model of Eq. 5, Divided Attention performs only (N+F+2)
comparisons per patch. Our experiments demonstrate that
this space-time factorization is not only more efficient but it
also leads to improved classification accuracy.
We have also experimented with a “Sparse Local Global”
(L+G) and an “Axial” (T+W+H) attention models. Their
architectures are illustrated in Fig. 1, while Fig. 2 shows
the patches considered for attention by these models. For
each patch (p;t), (L+G) first computes a local attention by
considering the neighboring FH=2W=2patches and
then calculates a sparse global attention over the entire clip
using a stride of 2 patches along the temporal dimension and
also the two spatial dimensions. Thus, it can be viewed as a
faster approximation of full spatiotemporal attention using a
local-global decomposition and a sparsity pattern, similar to
that used in (Child et al., 2019). Finally, “Axial” attention
decomposes the attention computation in three distinct steps:
over time, width and height. A decomposed attention over
the two spatial axes of the image was proposed in (Ho
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
252
|
decomposes the attention computation in three distinct steps:
over time, width and height. A decomposed attention over
the two spatial axes of the image was proposed in (Ho
et al., 2019; Huang et al., 2019; Wang et al., 2020b) and
our (T+W+H) adds a third dimension (time) for the case
of video. All these models are implemented by learning
distinct query/key/value matrices for each attention step.
4. Experiments
We evaluate TimeSformer on four popular action recognition datasets: Kinetics-400 (Carreira & Zisserman, 2017),
Kinetics-600 (Carreira et al., 2018), Something-SomethingV2 (Goyal et al., 2017b), and Diving-48 (Li et al., 2018). We
adopt the “Base” ViT architecture (Dosovitskiy et al., 2020)
pretrained on either ImageNet-1K or ImageNet-21K (Deng
et al., 2009), as specified for each experiment. Unless differently indicated, we use clips of size 8224224, with
frames sampled at a rate of 1=32. The patch size is 1616
pixels. During inference, unless otherwise noted, we sample a single temporal clip in the middle of the video. We
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
253
|
pixels. During inference, unless otherwise noted, we sample a single temporal clip in the middle of the video. We
use3spatial crops (top-left, center, bottom-right) from the
temporal clip and obtain the final prediction by averaging
the scores for these 3crops.
8 32 64 96# of Input frames0510TFLOPsJoint Space-TimeDivided Space-Time224 336 448 560Spatial Crop (Px)0123TFLOPsJoint Space-TimeDivided Space-Time
Out of memory
Out of memoryFigure 3. We compare the video classification cost (in TFLOPs) of
Joint Space-Time versus Divided Space-Time attention. We plot
the number of TFLOPs as a function of spatial crop size in pixels
(left), and the number of input frames (right). As we increase the
spatial resolution (left), or the video length (right), our proposed divided space-time attention leads to dramatic computational savings
compared to the scheme of joint space-time attention.
4.1. Analysis of Self-Attention Schemes
For this first set of experiments we start from a ViT pretrained on ImageNet-21K. In Table 1, we present the results
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
254
|
For this first set of experiments we start from a ViT pretrained on ImageNet-21K. In Table 1, we present the results
obtained with TimeSformer for the five proposed space-time
attention schemes on Kinetics-400 (K400) and SomethingSomething-V2 (SSv2). First, we note that TimeSformer
with space-only attention (S) performs well on K400. This
is an interesting finding. Indeed, prior work (Sevilla-Lara
et al., 2021) has shown that on K400, spatial cues are more
important than temporal information in order to achieve
strong accuracy. Here, we show that it is possible to obtain
solid accuracy on K400 without any temporal modeling.
Note, however, that space-only attention performs poorly on
SSv2. This stresses the importance of temporal modeling
on this latter dataset.
Furthermore, we observe that divided space-time attention
achieves the best accuracy on both K400 and SSv2. This
makes sense because compared to joint space-time attention,
divided space-time attention has a larger learning capacity
(see Table 1) as it contains distinct learning parameters for
temporal attention and spatial attention.
In Figure 3, we also compare the computational cost of joint
space-time versus divided space-time attention when using
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
255
|
temporal attention and spatial attention.
In Figure 3, we also compare the computational cost of joint
space-time versus divided space-time attention when using
higher spatial resolution (left) and longer (right) videos. We
note that the scheme of divided space-time scales gracefully
under both of these settings. In contrast, the scheme of
joint space-time attention leads to a dramatically higher cost
when resolution or video length is increased. In practice,
joint space-time attention causes a GPU memory overflow
once the spatial frame resolution reaches 448pixels, or once
the number of frames is increased to 32and thus it is effectively not applicable to large frames or long videos. Thus,
despite a larger number of parameters, divided space-time
attention is more efficient than joint space-time attention
when operating on higher spatial resolution, or longer videos.
Thus, for all subsequent experiments we use a TimeSformer
constructed with divided space-time self-attention blocks.
Is Space-Time Attention All You Need for Video Understanding?
Model Pretrain K400 Training
Time (hours)K400
Acc.Inference
TFLOPsParams
I3D 8x8 R50 ImageNet-1K 444 71.0 1.11 28.0M
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
256
|
Acc.Inference
TFLOPsParams
I3D 8x8 R50 ImageNet-1K 444 71.0 1.11 28.0M
I3D 8x8 R50 ImageNet-1K 1440 73.4 1.11 28.0M
SlowFast R50 ImageNet-1K 448 70.0 1.97 34.6M
SlowFast R50 ImageNet-1K 3840 75.6 1.97 34.6M
SlowFast R50 N/A 6336 76.4 1.97 34.6M
TimeSformer ImageNet-1K 416 75.8 0.59 121.4M
TimeSformer ImageNet-21K 416 78.0 0.59 121.4M
Table 2. Comparing TimeSformer to SlowFast and I3D. We observe that TimeSformer has lower inference cost despite having
a larger number of parameters. Furthermore, the cost of training
TimeSformer on video data is much lower compared to SlowFast
and I3D, even when all models are pretrained on ImageNet-1K.
4.2. Comparison to 3D CNNs
In this subsection we perform an empirical study aimed
at understanding the distinguishing properties of TimeSformer compared to 3D convolutional architectures, which
have been the prominent approach to video understanding
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
257
|
In this subsection we perform an empirical study aimed
at understanding the distinguishing properties of TimeSformer compared to 3D convolutional architectures, which
have been the prominent approach to video understanding
in recent years. We focus our comparison on two 3D CNN
models: 1) SlowFast (Feichtenhofer et al., 2019b), which is
the state-of-the-art in video classification, and 2) I3D (Carreira & Zisserman, 2017), which has been shown to benefit
from image-based pretraining, similarly to our own model.
We present quantitative comparisons to these two networks
in Table 2 and highlight key observations below.
Model Capacity. From Table 2, we first observe that although TimeSformer has a large learning capacity (the number of parameters is 121:4M), it has low inference cost
(0:59in TFLOPs). In contrast, SlowFast 8x8 R50 has a
larger inference cost ( 1:97TFLOPs) despite containing only
34:6Mparameters. Similarly, I3D 8x8 R50 also has a larger
inference cost ( 1:11TFLOPs) despite containing fewer parameters ( 28:0M). This suggests that TimeSformer is better
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
258
|
inference cost ( 1:11TFLOPs) despite containing fewer parameters ( 28:0M). This suggests that TimeSformer is better
suited for settings that involve large-scale learning. In contrast, the large computational cost of modern 3D CNNs
makes it difficult to further increase their model capacity
while also maintaining efficiency.
Video Training Time. One significant advantage of ImageNet pretraining is that it enables very efficient training of
TimeSformer on video data. Conversely, state-of-the-art 3D
CNNs are much more expensive to train even if pretrained
on image datasets. In Table 2, we compare the video training time on Kinetics-400 (in Tesla V100 GPU hours) of
TimeSformer to that of SlowFast and I3D. Starting from a
ResNet50 pretrained on ImageNet-1K, SlowFast 88R50
requires 3 840 Tesla V100 GPU hours in order to reach an
accuracy of 75:6%on Kinetics-400. Training I3D, under
similar settings, requires 1 440 Tesla V100 GPU hours for a
73:4%accuracy. In contrast, TimeSformer, also pretrained
on ImageNet-1K, only requires 416Tesla V100 GPU hours
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
259
|
73:4%accuracy. In contrast, TimeSformer, also pretrained
on ImageNet-1K, only requires 416Tesla V100 GPU hours
to achieve a higher 75:8%accuracy (see Table 2). Furthermore, if we constrain SlowFast to be trained under a
somewhat similar computational budget as TimeSformerMethod Pretraining K400 SSv2
TimeSformer ImageNet-1K 75.8 59.5
TimeSformer ImageNet-21K 78.0 59.5
TimeSformer-HR ImageNet-1K 77.8 62.2
TimeSformer-HR ImageNet-21K 79.7 62.5
TimeSformer-L ImageNet-1K 78.1 62.4
TimeSformer-L ImageNet-21K 80.7 62.3
Table 3. Comparing the effectiveness of ImageNet-1K and
ImageNet-21K pretraining on Kinetics-400 (K400) and SomethingSomething-V2 (SSv2). On K400, ImageNet-21K pretraining leads
consistently to a better performance compared to ImageNet-1K pretraining. On SSv2, ImageNet-1K and ImageNet-21K pretrainings
lead to similar accuracy.
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
260
|
lead to similar accuracy.
(i.e.,448GPU hours), its accuracy drops to 70:0%. Similarly, training I3D using a similar computational budget (i.e.,
444GPU hours) leads to a lower accuracy of 71:0%. This
highlights the fact that some of the latest 3D CNNs (Feichtenhofer et al., 2019b; Feichtenhofer, 2020) require a very
long optimization schedule to achieve good performance
(even when using ImageNet pretraining). In contrast, TimeSformer provides a more efficient alternative to labs that do
not have access to hundreds of GPUs.
The Importance of Pretraining. Due to a large number
of parameters, training our model from scratch is difficult.
Thus, before training TimeSformer on video data, we initialize it with weights learned from ImageNet. In contrast,
SlowFast can be learned on video data from scratch although
at the expense of a very high training cost (see Table 2). We
also attempted to train TimeSformer on Kinetics-400 directly, without any ImageNet pretraining. By using a longer
training schedule and more data augmentations, we found
it possible to train the model from scratch, albeit to a much
lower video level accuracy of 64:8%. Thus, based on these
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
261
|
training schedule and more data augmentations, we found
it possible to train the model from scratch, albeit to a much
lower video level accuracy of 64:8%. Thus, based on these
results, for all subsequent studies we continued to use ImageNet for pretraining (Deng et al., 2009)
In Table 3 we study the benefits of ImageNet-1K vs
ImageNet-21K pretraining on K400 and SSv2. For these
experiments, we use three variants of our model: (1) TimeSformer , which is the default version of our model operating
on8224224video clips, (2) TimeSformer-HR , a high
spatial resolution variant that operates on 16448448
video clips, and lastly (3) TimeSformer-L , a long-range
configuration of our model that operates on 96224224
video clips with frames sampled at a rate of 1=4.
Based on the results in Table 3, we observe that ImageNet21K pretraining is beneficial for K400, where it leads to
a consistently higher accuracy compared to ImageNet-1K
pretraining. On the other hand, on SSv2, we observe that
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
262
|
a consistently higher accuracy compared to ImageNet-1K
pretraining. On the other hand, on SSv2, we observe that
ImageNet-1K and ImageNet-21K pretrainings lead to similar accuracy. This makes sense as SSv2 requires complex
spatiotemporal reasoning, whereas K400 is biased more towards spatial scene information, and thus, it benefits more
from the features learned on the larger pretraining dataset.
Is Space-Time Attention All You Need for Video Understanding?
60K 120K 180K 240K# of Training Videos60657075AccuracyKinetics
TimeSformerSlowFastI3D42K 85K 127K 170K# of Training Videos354045505560AccuracySomething-Something-V2
TimeSformerSlowFastI3D
Figure 4. Accuracy on Kinetics-400 (K400), and SomethingSomething-V2 (SSv2) as a function of the number of training
videos. On K400, TimeSformer performs best in all cases. On
SSv2, which requires more complex temporal reasoning, TimeSformer outperforms the other models only when using enough
training videos. All models are pretrained on ImageNet-1K.
The Impact of Video-Data Scale. To understand the
effects of video-data scale on performance, we trained
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
263
|
training videos. All models are pretrained on ImageNet-1K.
The Impact of Video-Data Scale. To understand the
effects of video-data scale on performance, we trained
TimeSformer on different subsets of K400 and SSv2:
f25%;50%;75%;100%gof the full datasets. We show
these results in Figure 4, where we also compare our method
with SlowFast R50 (Feichtenhofer et al., 2019b), and I3D
R50 (Carreira & Zisserman, 2017) trained on the same subsets and using the same pretraining. Since we do not have
access to a ResNet pretrained on ImageNet-21K, we use
ImageNet-1K pretraining for all 3architectures.
The results of Figure 4 show that, on K400, TimeSformer
outperforms the other models for all training subsets. However, we observe a different trend on SSv2, where TimeSformer is the strongest model only when trained on 75% or
100% of the full data. This may be explained by the fact that
compared to K400, SSv2 requires learning more complex
temporal patterns, and thus more examples are needed by
TimeSformer to learn effectively those patterns.
4.3. Varying the Number of Tokens
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
264
|
temporal patterns, and thus more examples are needed by
TimeSformer to learn effectively those patterns.
4.3. Varying the Number of Tokens
The scalability of our model allows it to operate at higher
spatial resolution and on longer videos compared to most 3D
CNNs. We note that both of these aspects affect the length of
the sequence of tokens fed to the Transformer. Specifically,
increasing the spatial resolution results in a higher number
of patches (N) per frame. The number of input tokens is
also increased when using more frames. To investigate the
benefits, we conduct an empirical study where we separately
increase the number of tokens along each of these two axes.
We report the findings in Figure 5. We see that increasing
the spatial resolution (up to a certain point) leads to a boost
in performance. Similarly, we observe that increasing the
length of the input clip leads to consistent accuracy gains.
Due to GPU memory constraints, we are not able to test
our model on clips longer than 96frames. Still, we would
like to point out that using clips of 96frames is a significant
departure from current convolutional models, which are
typically limited to processing inputs of 8 32frames.
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
265
|
departure from current convolutional models, which are
typically limited to processing inputs of 8 32frames.
8 32 64 96Number of Input Frames65707580Clip Accuracy224 336 448 560Spatial Crop Side (Px)65707580Clip AccuracyFigure 5. Clip-level accuracy on Kinetics-400 as a function of spatial crop size in pixels (left), and the number of input frames (right).
Positional Embedding K400 SSv2
None 75.4 45.8
Space-only 77.8 52.5
Space-Time 78.0 59.5
Table 4. Ablation on positional embeddings. The version of TimeSformer using space-time positional embeddings yields the highest
accuracy on both Kinetics-400 and SSv2.
4.4. The Importance of Positional Embeddings
To investigate the importance of our learned spatiotemporal positional embeddings, we also conduct experiments
with a few variants of TimeSformer that use: (1) no positional embedding, (2) space-only positional embedding,
and (3) space-time positional embedding. We report these
results in Table 4. Based on these results, we observe
that the variant of our model that uses space-time positional embeddings produces the best accuracy on both
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
266
|
results in Table 4. Based on these results, we observe
that the variant of our model that uses space-time positional embeddings produces the best accuracy on both
Kinetics-400, and Something-Something-V2. Interestingly,
we also observe that using space-only positional embeddings leads to solid results on Kinetics-400, but much worse
results on Something-Something-V2. This makes sense as
Kinetics-400 is more spatially biased, whereas SomethingSomething-V2 requires complex temporal reasoning.
4.5. Comparison to the State-of-the-Art
Kinetics-400 & Kinetics-600. In Table 5 we present our
results on the validation set of K400. For these experiments,
we use TimeSformer pretrained on ImageNet-21K. In addition to the accuracy metrics, we also include inference
cost, given in TFLOPs. We note that whereas most previous
Method Top-1 Top-5 TFLOPs
R(2+1)D (Tran et al., 2018) 72.0 90.0 17.5
bLVNet (Fan et al., 2019) 73.5 91.2 0.84
TSM (Lin et al., 2019) 74.7 N/A N/A
S3D-G (Xie et al., 2018) 74.7 93.4 N/A
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
267
|
S3D-G (Xie et al., 2018) 74.7 93.4 N/A
Oct-I3D+NL (Chen et al., 2019) 75.7 N/A 0.84
D3D (Stroud et al., 2020) 75.9 N/A N/A
I3D+NL (Wang et al., 2018b) 77.7 93.3 10.8
ip-CSN-152 (Tran et al., 2019) 77.8 92.8 3.2
CorrNet (Wang et al., 2020a) 79.2 N/A 6.7
LGD-3D-101 (Qiu et al., 2019) 79.4 94.4 N/A
SlowFast (Feichtenhofer et al., 2019b) 79.8 93.9 7.0
X3D-XXL (Feichtenhofer, 2020) 80.4 94.6 5.8
TimeSformer 78.0 93.7 0.59
TimeSformer-HR 79.7 94.4 5.11
TimeSformer-L 80.7 94.7 7.14
Table 5. Video-level accuracy on Kinetics-400.
Is Space-Time Attention All You Need for Video Understanding?
Method Top-1 Top-5
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
268
|
Table 5. Video-level accuracy on Kinetics-400.
Is Space-Time Attention All You Need for Video Understanding?
Method Top-1 Top-5
I3D-R50+Cell (Wang et al., 2020c) 79.8 94.4
LGD-3D-101 (Qiu et al., 2019) 81.5 95.6
SlowFast (Feichtenhofer et al., 2019b) 81.8 95.1
X3D-XL (Feichtenhofer, 2020) 81.9 95.5
TimeSformer 79.1 94.4
TimeSformer-HR 81.8 95.8
TimeSformer-L 82.2 95.6
Table 6. Video-level accuracy on Kinetics-600.
1 3 5 10
# of Testing Clips65707580Accuracy
TimeSformer-L
TimeSformer
SlowFast-R101+NL
X3D-XL
Figure 6. Video-level accuracy on Kinetics-400 vs the number of
temporal clips used during inference. TimeSformer-L achieves
excellent accuracy using a small number of clips, which leads to
strong performance at low inference cost.
methods use 10temporal clips with 3spatial crops (for a total of 30space-time views) during inference, TimeSformer
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
269
|
strong performance at low inference cost.
methods use 10temporal clips with 3spatial crops (for a total of 30space-time views) during inference, TimeSformer
achieves solid accuracy with only 3views (3 spatial crops),
which reduces the inference cost. Our long-range variant,
TimeSformer-L achieves a top-1 accuracy of 80:7%. Furthermore, our default TimeSformer has the lowest inference
cost among recent state-of-the-art models. Yet, it still provides a solid accuracy of 78:0%, outperforming many more
costly models.
We also measured the actual inference runtime on 20Kvalidation videos of Kinetics-400 (using 8Tesla V100 GPUs).
Whereas SlowFast takes 14:88hours to complete the inference, TimeSformer, TimeSformer-HR, and TimeSformer-L
take36minutes, 1:06hours and 2:6hours, respectively.
Thus, even though SlowFast and TimeSformer-L have comparable cost in terms of TFLOPs, in practice the runtimes
of all our versions of TimeSformer are much lower.
In Table 6, we also present our results on Kinetics-600.
Just like on Kinetics-400, we observe that TimeSformer
performs well on this benchmark, outperforming all prior
methods.
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
2102.05095
|
270
|
Just like on Kinetics-400, we observe that TimeSformer
performs well on this benchmark, outperforming all prior
methods.
Finally, in Figure 6, we study the effect of using multiple temporal clips during inference (each with a single
spatial crop). We plot accuracy using K2f1;3;5;10g
temporal clips for testing. We compare our model against
X3D (Feichtenhofer, 2020), and SlowFast (Feichtenhofer
et al., 2019b). X3D and SlowFast require multiple ( 5)
clips to approach their top accuracy. Conversely, our longrange variant, TimeSformer-L, does not require multiple
clips to achieve its best performance, since it is able to spanMethod SSv2 Diving-48
SlowFast (Feichtenhofer et al., 2019b) 61.7 77.6
TSM (Lin et al., 2019) 63.4 N/A
STM (Jiang et al., 2019) 64.2 N/A
MSNet (Kwon et al., 2020) 64.7 N/A
TEA (Li et al., 2020b) 65.1 N/A
bLVNet (Fan et al., 2019) 65.2 N/A
TimeSformer 59.5 74.9
|
2102.05095
|
Is Space-Time Attention All You Need for Video Understanding?
|
We present a convolution-free approach to video classification built
exclusively on self-attention over space and time. Our method, named
"TimeSformer," adapts the standard Transformer architecture to video by
enabling spatiotemporal feature learning directly from a sequence of
frame-level patches. Our experimental study compares different self-attention
schemes and suggests that "divided attention," where temporal attention and
spatial attention are separately applied within each block, leads to the best
video classification accuracy among the design choices considered. Despite the
radically new design, TimeSformer achieves state-of-the-art results on several
action recognition benchmarks, including the best reported accuracy on
Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks,
our model is faster to train, it can achieve dramatically higher test
efficiency (at a small drop in accuracy), and it can also be applied to much
longer video clips (over one minute long). Code and models are available at:
https://github.com/facebookresearch/TimeSformer.
|
http://arxiv.org/pdf/2102.05095
|
[
"Gedas Bertasius",
"Heng Wang",
"Lorenzo Torresani"
] |
[
"cs.CV"
] |
Accepted to ICML 2021
| null |
cs.CV
|
20210209
|
20210609
|
[
{
"id": "2005.00928",
"title": "Quantifying Attention Flow in Transformers",
"authors": "Abnar, S. and Zuidema, W.",
"year": "2020"
},
{
"id": "1607.06450",
"title": "Layer Normalization",
"authors": "Lei Jimmy Ba, Jamie Ryan Kiros, Geoffrey E. Hinton",
"year": "2016"
},
{
"id": "1904.09925",
"title": "Attention Augmented Convolutional Networks",
"authors": "Bello, I., Zoph, B., Le, Q., Vaswani, A., Shlens, J.",
"year": "2019"
},
{
"id": "1912.04573",
"title": "Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation",
"authors": "Bertasius, G. and Torresani, L.",
"year": "2020"
},
{
"id": "2005.14165",
"title": "Language Models are Few-Shot Learners",
"authors": "Tom B. Brown, et al.",
"year": "2020"
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.