# PanopticFCN
**Fully Convolutional Networks for Panoptic Segmentation**
Yanwei Li, Hengshuang Zhao, Xiaojuan Qi, Liwei Wang, Zeming Li, Jian Sun, Jiaya Jia
[[`arXiv`](https://arxiv.org/pdf/2012.00720.pdf)] [[`BibTeX`](#CitingPanopticFCN)]
<div align="center">
<img src="docs/panoptic_fcn.png"/>
</div><br/>
This project provides an implementation for the CVPR 2021 Oral paper "[Fully Convolutional Networks for Panoptic Segmentation](https://arxiv.org/pdf/2012.00720.pdf)" based on [Detectron2](https://github.com/facebookresearch/detectron2). Panoptic FCN is a conceptually simple, strong, and efficient framework for panoptic segmentation, which represents and predicts foreground things and background stuff in a unified fully convolutional pipeline.
## Installation
This project is based on [Detectron2](https://github.com/facebookresearch/detectron2), which can be constructed as follows.
* Install Detectron2 following [the instructions](https://detectron2.readthedocs.io/tutorials/install.html).
* Setup the dataset following [the structure](https://github.com/facebookresearch/detectron2/blob/master/datasets/README.md).
* Copy this project to `/path/to/detectron2/projects/PanopticFCN`
## Training
To train a model with 8 GPUs, run:
```bash
cd /path/to/detectron2
python3 projects/PanopticFCN/train.py --config-file <config.yaml> --num-gpus 8
```
For example, to launch PanopticFCN training (1x schedule) with ResNet-50 backbone on 8 GPUs,
one should execute:
```bash
cd /path/to/detectron2
python3 projects/PanopticFCN/train.py --config-file projects/PanopticFCN/configs/PanopticFCN-R50-1x.yaml --num-gpus 8
```
## Evaluation
To evaluate a pre-trained model with 8 GPUs, run:
```bash
cd /path/to/detectron2
python3 projects/PanopticFCN/train.py --config-file <config.yaml> --num-gpus 8 --eval-only MODEL.WEIGHTS /path/to/model_checkpoint
```
## Results
We provide the results on COCO *val* set with pretrained models. *FPS* is measured on a single V100 GPU.
<table><tbody>
<!-- START TABLE -->
<!-- TABLE HEADER -->
<th valign="bottom">Method</th>
<th valign="bottom">Backbone</th>
<th valign="bottom">Sched</th>
<th valign="bottom">PQ</th>
<th valign="bottom">SQ</th>
<th valign="bottom">RQ</th>
<th valign="bottom">AP</th>
<th valign="bottom">mIoU</th>
<th valign="bottom">FPS</th>
<th valign="bottom">download</th>
<!-- TABLE BODY -->
<tr><td align="left">PanopticFCN</td>
<td align="center">R50</td>
<td align="center">1x</td>
<td align="center"> 41.1 </td>
<td align="center"> 79.8 </td>
<td align="center"> 49.9 </td>
<td align="center"> 32.2 </td>
<td align="center"> 41.5 </td>
<td align="center"> 12.4 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1tD1A5Zwbtri5OejlIz9MLKwzOzjtIMHQ/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1NeUO9EWtkZE0M5NrEpZ8uFqOX3vQg3Lx/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN-400</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 41.0 </td>
<td align="center"> 81.0 </td>
<td align="center"> 49.6 </td>
<td align="center"> 30.7 </td>
<td align="center"> 43.6 </td>
<td align="center"> 22.5 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1QBYMAznZDDX7A0Mnaq3euB23rTBzwUCf/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1QOwbA9KRIvDN8PKh10aCQhf1jpykKwbB/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN-512</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 42.3 </td>
<td align="center"> 81.1 </td>
<td align="center"> 51.2 </td>
<td align="center"> 32.4 </td>
<td align="center"> 43.2 </td>
<td align="center"> 19.8 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1QBYMAznZDDX7A0Mnaq3euB23rTBzwUCf/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1QOwbA9KRIvDN8PKh10aCQhf1jpykKwbB/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN-600</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 42.7 </td>
<td align="center"> 81.2 </td>
<td align="center"> 51.5 </td>
<td align="center"> 33.6 </td>
<td align="center"> 43.9 </td>
<td align="center"> 17.5 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1gIUxy1DJ_V91IwL5_jHQDMOIgHoWn_O1/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1OfbyJWIVfdGQ0C-JNUnXoocHXdILnIkf/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 43.6 </td>
<td align="center"> 81.4 </td>
<td align="center"> 52.5 </td>
<td align="center"> 34.4 </td>
<td align="center"> 43.6 </td>
<td align="center"> 12.8 </td>
<td align="center"> <a href="https://drive.google.com/file/d/18Re3keEkIiy7EVS-uFCNPBfT1BfT8Ng3/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1ACrIJ_AZCW3fD7jcipdya3-ixVBojnFO/view?usp=sharing">metrics</a></td>
</tr>
<tr><td align="left">PanopticFCN*</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 44.2 </td>
<td align="center"> 81.7 </td>
<td align="center"> 52.9 </td>
<td align="center"> 35.6 </td>
<td align="center"> 43.9 </td>
<td align="center"> 9.3 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1_VkJIhbQg9uqN49L3cDAW66zZKJE0fkI/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1uulb8PATBy1dF2VhlgQYQlo7gEdKRMK1/view?usp=sharing">metrics</a></td>
</tr>
</tbody></table>
A faster version is also provided with higher threshold but similar PQ results, which shares the same model with the corresponding normal one. This version could be suitable for you if the final panoptic results are taken into consideration only.
<table><tbody>
<!-- START TABLE -->
<!-- TABLE HEADER -->
<th valign="bottom">Method</th>
<th valign="bottom">Backbone</th>
<th valign="bottom">Sched</th>
<th valign="bottom">PQ</th>
<th valign="bottom">SQ</th>
<th valign="bottom">RQ</th>
<th valign="bottom">AP</th>
<th valign="bottom">mIoU</th>
<th valign="bottom">FPS</th>
<th valign="bottom">download</th>
<!-- TABLE BODY -->
<tr><td align="left">PanopticFCN</td>
<td align="center">R50</td>
<td align="center">1x</td>
<td align="center"> 41.1 </td>
<td align="center"> 79.8 </td>
<td align="center"> 49.9 </td>
<td align="center"> 30.2 </td>
<td align="center"> 41.4 </td>
<td align="center"> 13.6 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1tD1A5Zwbtri5OejlIz9MLKwzOzjtIMHQ/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1NeUO9EWtkZE0M5NrEpZ8uFqOX3vQg3Lx/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN-400</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 40.8 </td>
<td align="center"> 81.1 </td>
<td align="center"> 49.4 </td>
<td align="center"> 28.9 </td>
<td align="center"> 43.5 </td>
<td align="center"> 26.1 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1QBYMAznZDDX7A0Mnaq3euB23rTBzwUCf/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1QOwbA9KRIvDN8PKh10aCQhf1jpykKwbB/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN-512</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 42.3 </td>
<td align="center"> 81.1 </td>
<td align="center"> 51.2 </td>
<td align="center"> 30.7 </td>
<td align="center"> 43.2 </td>
<td align="center"> 22.0 </td>
<td align="center"> <a href="https://drive.google.com/file/d/1QBYMAznZDDX7A0Mnaq3euB23rTBzwUCf/view?usp=sharing">model</a> | <a href="https://drive.google.com/file/d/1QOwbA9KRIvDN8PKh10aCQhf1jpykKwbB/view?usp=sharing">metrics</a> </td>
</tr>
<tr><td align="left">PanopticFCN-600</td>
<td align="center">R50</td>
<td align="center">3x</td>
<td align="center"> 42.7 </td>
<td align="center"> 80.8 </td>
<td align="center"> 51.4 </td>
<td
没有合适的资源?快使用搜索试试~ 我知道了~
PanopticFCN:用于全景分割的全卷积网络(CVPR2021口服)
共26个文件
yaml:12个
py:11个
md:1个
需积分: 35 16 下载量 176 浏览量
2021-05-14
18:14:32
上传
评论 2
收藏 842KB ZIP 举报
温馨提示
全景光 全卷积网络用于全景分割 李彦伟,赵恒双,齐晓娟,王立伟,李泽明,孙健,贾佳亚 [ ] [ ] 该项目为基于的CVPR 2021口头论文“”提供了实现。 Panoptic FCN是一个概念上简单,强大且有效的全景图分割框架,它在统一的全卷积流水线中表示和预测前景事物和背景事物。 安装 该项目基于 ,它可以按以下方式构造。 按照安装Detectron2。 按照设置数据集。 将此项目复制到/path/to/detectron2/projects/PanopticFCN 训练 要使用8个GPU训练模型,请运行: cd /path/to/detectron2 python3 projects/PanopticFCN/train.py --config-file < config> --num-gpus 8 例如,要在8个GPU上使用ResNet-50主干网启动Pa
资源详情
资源评论
资源推荐
收起资源包目录
PanopticFCN-main.zip (26个子文件)
PanopticFCN-main
train.py 4KB
configs
PanopticFCN-R50-512-3x-FAST.yaml 1KB
PanopticFCN-R50-600-3x.yaml 1KB
PanopticFCN-R50-600-3x-FAST.yaml 1KB
PanopticFCN-R50-400-3x.yaml 1KB
PanopticFCN-R50-1x.yaml 1023B
PanopticFCN-R50-512-3x.yaml 1KB
PanopticFCN-R50-400-3x-FAST.yaml 1KB
PanopticFCN-R50-3x-FAST.yaml 1KB
PanopticFCN-R50-3x.yaml 1024B
PanopticFCN-Star-R50-3x-FAST.yaml 1KB
PanopticFCN-Star-R50-3x.yaml 1KB
PanopticFCN-R50-1x-FAST.yaml 1KB
LICENSE 11KB
README.md 10KB
panopticfcn
backbone_utils.py 4KB
config.py 3KB
panoptic_seg.py 26KB
utils.py 2KB
loss.py 4KB
__init__.py 126B
deform_conv_with_off.py 2KB
gt_generate.py 15KB
build_solver.py 3KB
head.py 11KB
docs
panoptic_fcn.png 837KB
共 26 条
- 1
WebWitch
- 粉丝: 21
- 资源: 4586
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
最新资源
- 筷手引流工具.apk
- 论文(最终)_20240430235101.pdf
- 基于python编写的Keras深度学习框架开发,利用卷积神经网络CNN,快速识别图片并进行分类
- 最全空间计量实证方法(空间杜宾模型和检验以及结果解释文档).txt
- 5uonly.apk
- 蓝桥杯Python组的历年真题
- 2023-04-06-项目笔记 - 第一百十九阶段 - 4.4.2.117全局变量的作用域-117 -2024.04.30
- 2023-04-06-项目笔记 - 第一百十九阶段 - 4.4.2.117全局变量的作用域-117 -2024.04.30
- 前端开发技术实验报告:内含4四实验&实验报告
- Highlight Plus v20.0.1
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0