Home

Awesome

CenterMask2

[CenterMask(original code)][vovnet-detectron2][arxiv] [BibTeX]

CenterMask2 is an upgraded implementation on top of detectron2 beyond original CenterMask based on maskrcnn-benchmark.

CenterMask : Real-Time Anchor-Free Instance Segmentation (CVPR 2020)<br> Youngwan Lee and Jongyoul Park<br> Electronics and Telecommunications Research Institute (ETRI)<br> pre-print : https://arxiv.org/abs/1911.06667

<div align="center"> <img src="https://dl.dropbox.com/s/yg9zr1tvljoeuyi/architecture.png" width="850px" /> </div>

Highlights

Updates

Results on COCO val

Note

We measure the inference time of all models with batch size 1 on the same V100 GPU machine.

CenterMask

MethodBackbonelr schedinference timemask APbox APdownload
Mask R-CNN (detectron2)R-503x0.05537.241.0<a href="https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl">model</a> | <a href="https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/metrics.json">metrics</a>
Mask R-CNN (detectron2)V2-393x0.05239.343.8<a href="https://dl.dropbox.com/s/dkto39ececze6l4/faster_V_39_eSE_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/dx9qz1dn65ccrwd/faster_V_39_eSE_ms_3x_metrics.json">metrics</a>
CenterMask (maskrcnn-benchmark)V2-393x0.07038.543.5link
CenterMask2V2-393x0.05039.744.2<a href="https://dl.dropbox.com/s/tczecsdxt10uai5/centermask2-V-39-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/rhoo6vkvh7rjdf9/centermask2-V-39-eSE-FPN-ms-3x_metrics.json">metrics</a>
Mask R-CNN (detectron2)R-1013x0.07038.642.9<a href="https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_101_FPN_3x/138205316/model_final_a3ec72.pkl">model</a> | <a href="https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_101_FPN_3x/138205316/metrics.json">metrics</a>
Mask R-CNN (detectron2)V2-573x0.05839.744.2<a href="https://dl.dropbox.com/s/c7mb1mq10eo4pzk/faster_V_57_eSE_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/3tsn218zzmuhyo8/faster_V_57_eSE_metrics.json">metrics</a>
CenterMask (maskrcnn-benchmark)V2-573x0.07639.444.6link
CenterMask2V2-573x0.05840.545.1<a href="https://dl.dropbox.com/s/lw8nxajv1tim8gr/centermask2-V-57-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/x7r5ys3c81ldgq0/centermask2-V-57-eSE-FPN-ms-3x_metrics.json">metrics</a>
Mask R-CNN (detectron2)X-1013x0.12939.544.3<a href="https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_X_101_32x8d_FPN_3x/139653917/model_final_2d9806.pkl">model</a> | <a href="https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_X_101_32x8d_FPN_3x/139653917/metrics.json">metrics</a>
Mask R-CNN (detectron2)V2-993x0.07640.344.9<a href="https://dl.dropbox.com/s/v64mknwzfpmfcdh/faster_V_99_eSE_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/zvaz9s8gvq2mhrd/faster_V_99_eSE_ms_3x_metrics.json">metrics</a>
CenterMask (maskrcnn-benchmark)V2-993x0.10640.245.6link
CenterMask2V2-993x0.07741.446.0<a href="https://dl.dropbox.com/s/c6n79x83xkdowqc/centermask2-V-99-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/jdzgmdatit00hq5/centermask2-V-99-eSE-FPN-ms-3x_metrics.json">metrics</a>
CenterMask2 (TTA)V2-993x-42.548.6<a href="https://dl.dropbox.com/s/c6n79x83xkdowqc/centermask2-V-99-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/jdzgmdatit00hq5/centermask2-V-99-eSE-FPN-ms-3x_metrics.json">metrics</a>

CenterMask-Lite

MethodBackbonelr schedinference timemask APbox APdownload
YOLACT550R-504x0.02328.230.3link
CenterMask (maskrcnn-benchmark)V-194x0.02332.435.9link
CenterMask2-LiteV-194x0.02332.835.9<a href="https://dl.dropbox.com/s/dret2ap7djty7mp/centermask2-lite-V-19-eSE-FPN-ms-4x.pth">model</a> | <a href="https://dl.dropbox.com/s/zsta7azy87a833u/centermask2-lite-V-19-eSE-FPN-ms-4x-metrics.json">metrics</a>
YOLACT550R-1014x0.03028.230.3link
YOLACT550++R-504x0.02934.1-link
YOLACT550++R-1014x0.03634.6-link
CenterMask (maskrcnn-benchmark)V-394x0.02736.340.7link
CenterMask2-LiteV-394x0.02836.740.9<a href="https://dl.dropbox.com/s/uwc0ypa1jvco2bi/centermask2-lite-V-39-eSE-FPN-ms-4x.pth">model</a> | <a href="https://dl.dropbox.com/s/aoa6y3i3el4edbk/centermask2-lite-V-39-eSE-FPN-ms-4x-metrics.json">metrics</a>

Lightweight VoVNet backbone

MethodBackboneParam.lr schedinference timemask APbox APdownload
CenterMask2-LiteMobileNetV23.5M4x0.02127.229.8<a href="https://dl.dropbox.com/s/8omou546f0n78nj/centermask_lite_Mv2_ms_4x.pth">model</a> | <a href="https://dl.dropbox.com/s/2jlcwy30eq72w47/centermask_lite_Mv2_ms_4x_metrics.json">metrics</a>
CenterMask2-LiteV-1911.2M4x0.02332.835.9<a href="https://dl.dropbox.com/s/dret2ap7djty7mp/centermask2-lite-V-19-eSE-FPN-ms-4x.pth">model</a> | <a href="https://dl.dropbox.com/s/zsta7azy87a833u/centermask2-lite-V-19-eSE-FPN-ms-4x-metrics.json">metrics</a>
CenterMask2-LiteV-19-Slim3.1M4x0.02129.832.5<a href="https://dl.dropbox.com/s/o2n1ifl0zkbv16x/centermask-lite-V-19-eSE-slim-FPN-ms-4x.pth">model</a> | <a href="https://dl.dropbox.com/s/8y71oz0kxwqk7go/centermask-lite-V-19-eSE-slim-FPN-ms-4x-metrics.json?dl=0">metrics</a>
CenterMask2-LiteV-19Slim-DW1.8M4x0.02027.129.5<a href="https://dl.dropbox.com/s/vsvhwtqm6ko1c7m/centermask-lite-V-19-eSE-slim-dw-FPN-ms-4x.pth">model</a> | <a href="https://dl.dropbox.com/s/q4idjnsgvo151zx/centermask-lite-V-19-eSE-slim-dw-FPN-ms-4x-metrics.json">metrics</a>

Deformable VoVNet Backbone

MethodBackbonelr schedinference timemask APbox APdownload
CenterMask2V2-393x0.05039.744.2<a href="https://dl.dropbox.com/s/tczecsdxt10uai5/centermask2-V-39-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/rhoo6vkvh7rjdf9/centermask2-V-39-eSE-FPN-ms-3x_metrics.json">metrics</a>
CenterMask2V2-39-DCN3x0.06140.345.1<a href="https://dl.dropbox.com/s/zmps03vghzirk7v/centermask-V-39-eSE-dcn-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/aj1mr8m32z11zbw/centermask-V-39-eSE-dcn-FPN-ms-3x-metrics.json">metrics</a>
CenterMask2V2-573x0.05840.545.1<a href="https://dl.dropbox.com/s/lw8nxajv1tim8gr/centermask2-V-57-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/x7r5ys3c81ldgq0/centermask2-V-57-eSE-FPN-ms-3x_metrics.json">metrics</a>
CenterMask2V2-57-DCN3x0.07140.945.5<a href="https://dl.dropbox.com/s/1f64azqyd2ot6qq/centermask-V-57-eSE-dcn-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/b3zpguko137r6eh/centermask-V-57-eSE-dcn-FPN-ms-3x-metrics.json">metrics</a>
CenterMask2V2-993x0.07741.446.0<a href="https://dl.dropbox.com/s/c6n79x83xkdowqc/centermask2-V-99-eSE-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/jdzgmdatit00hq5/centermask2-V-99-eSE-FPN-ms-3x_metrics.json">metrics</a>
CenterMask2V2-99-DCN3x0.11042.046.9<a href="https://dl.dropbox.com/s/atuph90nzm7s8x8/centermask-V-99-eSE-dcn-FPN-ms-3x.pth">model</a> | <a href="https://dl.dropbox.com/s/82ulexlivy19cve/centermask-V-99-eSE-dcn-FPN-ms-3x-metrics.json">metrics</a>

Panoptic-CenterMask

MethodBackbonelr schedinference timemask APbox APPQdownload
Panoptic-FPNR-503x0.06340.036.541.5<a href="https://dl.fbaipublicfiles.com/detectron2/COCO-PanopticSegmentation/panoptic_fpn_R_50_3x/139514569/model_final_c10459.pkl">model</a> | <a href="https://dl.fbaipublicfiles.com/detectron2/COCO-PanopticSegmentation/panoptic_fpn_R_50_3x/139514569/metrics.json">metrics</a>
Panoptic-CenterMaskR-503x0.06341.437.342.0<a href="https://dl.dropbox.com/s/vxe51cdeprao94j/panoptic_centermask_R_50_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/dfddgx6rnw1zr4l/panoptic_centermask_R_50_ms_3x_metrics.json">metrics</a>
Panoptic-FPNV-393x0.06342.838.543.4<a href="https://dl.dropbox.com/s/fnr9r4arv0cbfbf/panoptic_V_39_eSE_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/vftfukrjuu7w1ao/panoptic_V_39_eSE_3x_metrics.json">metrics</a>
Panoptic-CenterMaskV-393x0.06643.439.043.7<a href="https://dl.dropbox.com/s/49ig16ailra1f4t/panoptic_centermask_V_39_eSE_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/wy4mn8n513k0um5/panoptic_centermask_V_39_eSE_ms_3x_metrics.json">metrics</a>
Panoptic-FPNR-1013x0.07842.438.543.0<a href="https://dl.fbaipublicfiles.com/detectron2/COCO-PanopticSegmentation/panoptic_fpn_R_101_3x/139514519/model_final_cafdb1.pkl">model</a> | <a href="https://dl.fbaipublicfiles.com/detectron2/COCO-PanopticSegmentation/panoptic_fpn_R_101_3x/139514519/metrics.json">metrics</a>
Panoptic-CenterMaskR-1013x0.07643.539.043.6<a href="https://dl.dropbox.com/s/y5stg3qx72gff5o/panoptic_centermask_R_101_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/ojljt0obp8vnr8s/panoptic_centermask_R_101_ms_3x_metrics.json">metrics</a>
Panoptic-FPNV-573x0.07043.439.244.3<a href="https://www.dropbox.com/s/zhoqx5rvc0jj0oa/panoptic_V_57_eSE_3x.pth?dl=1">model</a> | <a href="https://dl.dropbox.com/s/20hwrmru15dilre/panoptic_V_57_eSE_3x_metrics.json">metrics</a>
Panoptic-CenterMaskV-573x0.07143.939.644.5<a href="https://dl.dropbox.com/s/kqukww4y7tbgbrh/panoptic_centermask_V_57_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/4asto3b4iya74ak/panoptic_centermask_V_57_ms_3x_metrics.json">metrics</a>
Panoptic-CenterMaskV-993x0.09145.140.645.4<a href="https://dl.dropbox.com/s/pr6a3inpasn7qlz/panoptic_centermask_V_99_ms_3x.pth">model</a> | <a href="https://dl.dropbox.com/s/00e8x0riplme7pm/panoptic_centermask_V_99_ms_3x_metrics.json">metrics</a>

Installation

All you need to use centermask2 is detectron2. It's easy!
you just install detectron2 following INSTALL.md.
Prepare for coco dataset following this instruction.

Training

ImageNet Pretrained Models

We provide backbone weights pretrained on ImageNet-1k dataset for detectron2.

To train a model, run

cd centermask2
python train_net.py --config-file "configs/<config.yaml>"

For example, to launch CenterMask training with VoVNetV2-39 backbone on 8 GPUs, one should execute:

cd centermask2
python train_net.py --config-file "configs/centermask/centermask_V_39_eSE_FPN_ms_3x.yaml" --num-gpus 8

Evaluation

Model evaluation can be done similarly:

cd centermask2
wget https://dl.dropbox.com/s/tczecsdxt10uai5/centermask2-V-39-eSE-FPN-ms-3x.pth
python train_net.py --config-file "configs/centermask/centermask_V_39_eSE_FPN_ms_3x.yaml" --num-gpus 1 --eval-only MODEL.WEIGHTS centermask2-V-39-eSE-FPN-ms-3x.pth

TODO

<a name="CitingCenterMask"></a>Citing CenterMask

If you use VoVNet, please use the following BibTeX entry.

@inproceedings{lee2019energy,
  title = {An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection},
  author = {Lee, Youngwan and Hwang, Joong-won and Lee, Sangrok and Bae, Yuseok and Park, Jongyoul},
  booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
  year = {2019}
}

@inproceedings{lee2020centermask,
  title={CenterMask: Real-Time Anchor-Free Instance Segmentation},
  author={Lee, Youngwan and Park, Jongyoul},
  booktitle={CVPR},
  year={2020}
}

Special Thanks to

mask scoring for detectron2 by Sangrok Lee
FCOS_for_detectron2 by AdeliDet team.