# SOLOv2: Dynamic and Fast Instance Segmentation
> [**SOLOv2: Dynamic and Fast Instance Segmentation**](https://arxiv.org/abs/2003.10152),
> Xinlong Wang, Rufeng Zhang, Tao Kong, Lei Li, Chunhua Shen
> In: Proc. Advances in Neural Information Processing Systems (NeurIPS), 2020
> *arXiv preprint ([arXiv 2003.10152](https://arxiv.org/abs/2003.10152))*
# Installation & Quick Start
First, follow the [default instruction](../../README.md#Installation) to install the project and [datasets/README.md](https://github.com/facebookresearch/detectron2/blob/master/datasets/README.md)
set up the datasets (e.g., MS-COCO).
For demo, run the following command lines:
```
wget https://cloudstor.aarnet.edu.au/plus/s/chF3VKQT4RDoEqC/download -O SOLOv2_R50_3x.pth
python demo/demo.py \
--config-file configs/SOLOv2/R50_3x.yaml \
--input input1.jpg input2.jpg \
--opts MODEL.WEIGHTS SOLOv2_R50_3x.pth
```
For training on COCO, run:
```
OMP_NUM_THREADS=1 python tools/train_net.py \
--config-file configs/SOLOv2/R50_3x.yaml \
--num-gpus 8 \
OUTPUT_DIR training_dir/SOLOv2_R50_3x
```
For evaluation on COCO, run:
```
OMP_NUM_THREADS=1 python tools/train_net.py \
--config-file configs/SOLOv2/R50_3x.yaml \
--eval-only \
--num-gpus 8 \
OUTPUT_DIR training_dir/SOLOv2_R50_3x \
MODEL.WEIGHTS training_dir/SOLOv2_R50_3x/model_final.pth
```
## Models
### COCO Instance Segmentation Baselines with SOLOv2
Name | inf. time | train. time | Mem | box AP | mask AP | download
--- |:---:|:---:|:---:|:---:|:---:|:---:
[SOLOv2_R50_3x](R50_3x.yaml) | 47ms | ~25h(36 epochs) | 3.7GB | - | 37.6 | [model](https://cloudstor.aarnet.edu.au/plus/s/chF3VKQT4RDoEqC/download)
[SOLOv2_R101_3x](R101_3x.yaml) | 61ms | ~30h(36 epochs) | 4.7GB | - | 39.0 | [model](https://cloudstor.aarnet.edu.au/plus/s/9w7b3sjaXvqYQEQ)
*Disclaimer:*
- All models are trained with multi-scale data augmentation.
- Inference time is measured on a single V100 GPU. Training time is measured on 8 V100 GPUs.
- This is a reimplementation. Thus, the numbers are slightly different from our original paper (within 0.3% in mask AP).
- The implementation on mmdetection is available at [https://github.com/WXinlong/SOLO](https://github.com/WXinlong/SOLO).
# Citations
Please consider citing our papers in your publications if the project helps your research. BibTeX reference is as follows.
```BibTeX
@inproceedings{wang2020solo,
title = {{SOLO}: Segmenting Objects by Locations},
author = {Wang, Xinlong and Kong, Tao and Shen, Chunhua and Jiang, Yuning and Li, Lei},
booktitle = {Proc. Eur. Conf. Computer Vision (ECCV)},
year = {2020}
}
```
```BibTeX
@inproceedings{wang2020solov2,
title = {{SOLOv2}: Dynamic and Fast Instance Segmentation},
author = {Wang, Xinlong and Zhang, Rufeng and Kong, Tao and Li, Lei and Shen, Chunhua},
booktitle = {Proc. Advances in Neural Information Processing Systems (NeurIPS)},
year = {2020}
}
```
没有合适的资源?快使用搜索试试~ 我知道了~
温馨提示
基于python开发的工业场景的脱轨检测.+源码,适合毕业设计、课程设计、项目开发。项目源码已经过严格测试,可以放心参考并在此基础上延申使用~ 基于python开发的工业场景的脱轨检测.+源码,适合毕业设计、课程设计、项目开发。项目源码已经过严格测试,可以放心参考并在此基础上延申使用 项目简介: 这个项目基于Adelaidet,但为了简单起见,我们删除了所有与SOLOv2无关的文件。本回购的目的是: 培训城市景观&&自定义数据集; 尝试不同的主干,例如分段; 在不同的SxS设置上进行实验,以提高小物体的精度; ONNX导出; TensorRT部署;
资源推荐
资源详情
资源评论
收起资源包目录
工业场景的脱轨检测.zip (96个子文件)
outOfRailWay-main
zhibiao.png 68KB
tools
remove_optim_from_ckpt.py 589B
convert_fcos_weight.py 2KB
train_taco.py 11KB
train_microship.py 9KB
compute_flops.py 904B
visualize_data.py 4KB
train_my.py 9KB
rename_blendmask.py 1KB
train_net.py 7KB
log.md 1KB
tuoguiresult.jpg 83KB
setup.py 3KB
result2.png 18.22MB
data
microship 82B
b1.jpg 63KB
b2.jpg 28KB
run1.mp4 1.1MB
run2.mp4 1.22MB
taco 58B
custom_data 77B
coco 49B
MODEL_ZOO.md 3KB
LICENSE 1KB
configs
SOLOv2
Base-SOLOv2.yaml 512B
custom_dataset
Base-SOLOv2.yaml 506B
SOLOv2_lite.yaml 802B
R50_lite_3x.yaml 171B
R50_3x.yaml 171B
microship_dataset
Base-SOLOv2.yaml 511B
SOLOv2_lite.yaml 810B
R50_lite_3x.yaml 171B
R101_3x.yaml 173B
taco
Base-SOLOv2.yaml 511B
SOLOv2_lite.yaml 800B
R50_lite_3x.yaml 171B
R50_3x.yaml 171B
README.md 3KB
coco
SOLOv2_lite.yaml 808B
R50_lite_3x.yaml 171B
demo
demo_fast.py 6KB
predictor.py 9KB
predictorBak.py 9KB
demo.py 6KB
demoBak.py 6KB
datasets
gen_coco_person.py 3KB
prepare_thing_sem_from_lvis.py 3KB
prepare_thing_sem_from_instance.py 4KB
README.md 2KB
xingneng.png 50KB
adet
checkpoint
__init__.py 78B
adet_checkpoint.py 2KB
layers
__init__.py 364B
ml_nms.py 834B
csrc
cuda_version.cu 122B
vision.cpp 1KB
DefROIAlign
DefROIAlign_cuda.cu 15KB
DefROIAlign.h 3KB
ml_nms
ml_nms.cu 5KB
ml_nms.h 787B
gcn.py 3KB
iou_loss.py 2KB
conv_with_kaiming_uniform.py 2KB
deform_conv.py 4KB
def_roi_align.py 3KB
naive_group_norm.py 3KB
data
__init__.py 158B
augmentation.py 4KB
dataset_mapper.py 8KB
builtin.py 2KB
datasets
text.py 8KB
detection_utils.py 3KB
utils
measures.py 7KB
comm.py 1KB
visualizer.py 3KB
config
__init__.py 58B
defaults.py 11KB
config.py 225B
modeling
__init__.py 336B
solov2
utils.py 7KB
__init__.py 27B
loss.py 4KB
solov2.py 42KB
one_stage_detector.py 7KB
backbone
__init__.py 264B
resnet_lpf.py 12KB
vovnet.py 11KB
fpn.py 3KB
bifpn.py 15KB
dla.py 15KB
lpf.py 4KB
mobilenet.py 5KB
resnet_interval.py 5KB
result1.png 18.22MB
README.md 592B
detectron2
1.txt 8B
共 96 条
- 1
资源评论
梦回阑珊
- 粉丝: 2952
- 资源: 798
下载权益
C知道特权
VIP文章
课程特权
开通VIP
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
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