68 lines
15 KiB
Markdown
68 lines
15 KiB
Markdown
# NonLocal Net
|
|
|
|
[Non-local Neural Networks](https://arxiv.org/abs/1711.07971)
|
|
|
|
## Introduction
|
|
|
|
<!-- [ALGORITHM] -->
|
|
|
|
<a href="https://github.com/facebookresearch/video-nonlocal-net">Official Repo</a>
|
|
|
|
<a href="https://github.com/open-mmlab/mmsegmentation/blob/v0.17.0/mmseg/models/decode_heads/nl_head.py#L10">Code Snippet</a>
|
|
|
|
## Abstract
|
|
|
|
<!-- [ABSTRACT] -->
|
|
|
|
Both convolutional and recurrent operations are building blocks that process one local neighborhood at a time. In this paper, we present non-local operations as a generic family of building blocks for capturing long-range dependencies. Inspired by the classical non-local means method in computer vision, our non-local operation computes the response at a position as a weighted sum of the features at all positions. This building block can be plugged into many computer vision architectures. On the task of video classification, even without any bells and whistles, our non-local models can compete or outperform current competition winners on both Kinetics and Charades datasets. In static image recognition, our non-local models improve object detection/segmentation and pose estimation on the COCO suite of tasks. Code is available at [this https URL](https://github.com/facebookresearch/video-nonlocal-net).
|
|
|
|
<!-- [IMAGE] -->
|
|
<div align=center>
|
|
<img src="https://user-images.githubusercontent.com/24582831/142902128-17e29678-bf12-4ff4-b3d6-a39b47dfd253.png" width="50%"/>
|
|
</div>
|
|
|
|
## Citation
|
|
|
|
```bibtex
|
|
@inproceedings{wang2018non,
|
|
title={Non-local neural networks},
|
|
author={Wang, Xiaolong and Girshick, Ross and Gupta, Abhinav and He, Kaiming},
|
|
booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
|
|
pages={7794--7803},
|
|
year={2018}
|
|
}
|
|
```
|
|
|
|
## Results and models
|
|
|
|
### Cityscapes
|
|
|
|
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | config | download |
|
|
| -------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------- | ----------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
|
|
| NonLocalNet | R-50-D8 | 512x1024 | 40000 | 7.4 | 2.72 | 78.24 | - | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_512x1024_40k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x1024_40k_cityscapes/nonlocal_r50-d8_512x1024_40k_cityscapes_20200605_210748-c75e81e3.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x1024_40k_cityscapes/nonlocal_r50-d8_512x1024_40k_cityscapes_20200605_210748.log.json) |
|
|
| NonLocalNet | R-101-D8 | 512x1024 | 40000 | 10.9 | 1.95 | 78.66 | - | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_512x1024_40k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x1024_40k_cityscapes/nonlocal_r101-d8_512x1024_40k_cityscapes_20200605_210748-d63729fa.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x1024_40k_cityscapes/nonlocal_r101-d8_512x1024_40k_cityscapes_20200605_210748.log.json) |
|
|
| NonLocalNet | R-50-D8 | 769x769 | 40000 | 8.9 | 1.52 | 78.33 | 79.92 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_769x769_40k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_769x769_40k_cityscapes/nonlocal_r50-d8_769x769_40k_cityscapes_20200530_045243-82ef6749.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_769x769_40k_cityscapes/nonlocal_r50-d8_769x769_40k_cityscapes_20200530_045243.log.json) |
|
|
| NonLocalNet | R-101-D8 | 769x769 | 40000 | 12.8 | 1.05 | 78.57 | 80.29 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_769x769_40k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_769x769_40k_cityscapes/nonlocal_r101-d8_769x769_40k_cityscapes_20200530_045348-8fe9a9dc.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_769x769_40k_cityscapes/nonlocal_r101-d8_769x769_40k_cityscapes_20200530_045348.log.json) |
|
|
| NonLocalNet | R-50-D8 | 512x1024 | 80000 | - | - | 78.01 | - | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_512x1024_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x1024_80k_cityscapes/nonlocal_r50-d8_512x1024_80k_cityscapes_20200607_193518-d6839fae.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x1024_80k_cityscapes/nonlocal_r50-d8_512x1024_80k_cityscapes_20200607_193518.log.json) |
|
|
| NonLocalNet | R-101-D8 | 512x1024 | 80000 | - | - | 78.93 | - | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_512x1024_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x1024_80k_cityscapes/nonlocal_r101-d8_512x1024_80k_cityscapes_20200607_183411-32700183.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x1024_80k_cityscapes/nonlocal_r101-d8_512x1024_80k_cityscapes_20200607_183411.log.json) |
|
|
| NonLocalNet | R-50-D8 | 769x769 | 80000 | - | - | 79.05 | 80.68 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_769x769_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_769x769_80k_cityscapes/nonlocal_r50-d8_769x769_80k_cityscapes_20200607_193506-1f9792f6.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_769x769_80k_cityscapes/nonlocal_r50-d8_769x769_80k_cityscapes_20200607_193506.log.json) |
|
|
| NonLocalNet | R-101-D8 | 769x769 | 80000 | - | - | 79.40 | 80.85 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_769x769_80k_cityscapes.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_769x769_80k_cityscapes/nonlocal_r101-d8_769x769_80k_cityscapes_20200607_183428-0e1fa4f9.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_769x769_80k_cityscapes/nonlocal_r101-d8_769x769_80k_cityscapes_20200607_183428.log.json) |
|
|
|
|
### ADE20K
|
|
|
|
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | config | download |
|
|
| -------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | ------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
|
|
| NonLocalNet | R-50-D8 | 512x512 | 80000 | 9.1 | 21.37 | 40.75 | 42.05 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_80k_ade20k/nonlocal_r50-d8_512x512_80k_ade20k_20200615_015801-5ae0aa33.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_80k_ade20k/nonlocal_r50-d8_512x512_80k_ade20k_20200615_015801.log.json) |
|
|
| NonLocalNet | R-101-D8 | 512x512 | 80000 | 12.6 | 13.97 | 42.90 | 44.27 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_80k_ade20k/nonlocal_r101-d8_512x512_80k_ade20k_20200615_015758-24105919.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_80k_ade20k/nonlocal_r101-d8_512x512_80k_ade20k_20200615_015758.log.json) |
|
|
| NonLocalNet | R-50-D8 | 512x512 | 160000 | - | - | 42.03 | 43.04 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_160k_ade20k/nonlocal_r50-d8_512x512_160k_ade20k_20200616_005410-baef45e3.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_160k_ade20k/nonlocal_r50-d8_512x512_160k_ade20k_20200616_005410.log.json) |
|
|
| NonLocalNet | R-101-D8 | 512x512 | 160000 | - | - | 44.63 | 45.79 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_160k_ade20k/nonlocal_r101-d8_512x512_160k_ade20k_20210827_221502-7881aa1a.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_160k_ade20k/nonlocal_r101-d8_512x512_160k_ade20k_20210827_221502.log.json) |
|
|
|
|
### Pascal VOC 2012 + Aug
|
|
|
|
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | config | download |
|
|
| -------- | -------- | --------- | ------: | -------- | -------------- | ----: | ------------: | -------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
|
|
| NonLocalNet | R-50-D8 | 512x512 | 20000 | 6.4 | 21.21 | 76.20 | 77.12 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_512x512_20k_voc12aug.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_20k_voc12aug/nonlocal_r50-d8_512x512_20k_voc12aug_20200617_222613-07f2a57c.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_20k_voc12aug/nonlocal_r50-d8_512x512_20k_voc12aug_20200617_222613.log.json) |
|
|
| NonLocalNet | R-101-D8 | 512x512 | 20000 | 9.8 | 14.01 | 78.15 | 78.86 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_512x512_20k_voc12aug.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_20k_voc12aug/nonlocal_r101-d8_512x512_20k_voc12aug_20200617_222615-948c68ab.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_20k_voc12aug/nonlocal_r101-d8_512x512_20k_voc12aug_20200617_222615.log.json) |
|
|
| NonLocalNet | R-50-D8 | 512x512 | 40000 | - | - | 76.65 | 77.47 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r50-d8_512x512_40k_voc12aug.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_40k_voc12aug/nonlocal_r50-d8_512x512_40k_voc12aug_20200614_000028-0139d4a9.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r50-d8_512x512_40k_voc12aug/nonlocal_r50-d8_512x512_40k_voc12aug_20200614_000028.log.json) |
|
|
| NonLocalNet | R-101-D8 | 512x512 | 40000 | - | - | 78.27 | 79.12 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/nonlocal_net/nonlocal_r101-d8_512x512_40k_voc12aug.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_40k_voc12aug/nonlocal_r101-d8_512x512_40k_voc12aug_20200614_000028-7e5ff470.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/nonlocal_net/nonlocal_r101-d8_512x512_40k_voc12aug/nonlocal_r101-d8_512x512_40k_voc12aug_20200614_000028.log.json) |
|