Skip to content

Latest commit

 

History

History
46 lines (39 loc) · 8.59 KB

README.md

File metadata and controls

46 lines (39 loc) · 8.59 KB

Deformable Convolutional Networks

Introduction

@inproceedings{dai2017deformable,
  title={Deformable Convolutional Networks},
  author={Dai, Jifeng and Qi, Haozhi and Xiong, Yuwen and Li, Yi and Zhang, Guodong and Hu, Han and Wei, Yichen},
  booktitle={Proceedings of the IEEE international conference on computer vision},
  year={2017}
}

@article{zhu2018deformable,
  title={Deformable ConvNets v2: More Deformable, Better Results},
  author={Zhu, Xizhou and Hu, Han and Lin, Stephen and Dai, Jifeng},
  journal={arXiv preprint arXiv:1811.11168},
  year={2018}
}

Results and Models

Backbone Model Style Conv Pool Lr schd Mem (GB) Inf time (fps) box AP mask AP Download
R-50-FPN Faster pytorch dconv(c3-c5) - 1x 4.0 17.8 41.3 model | log
R-50-FPN Faster pytorch mdconv(c3-c5) - 1x 4.1 17.6 41.4 model | log
*R-50-FPN (dg=4) Faster pytorch mdconv(c3-c5) - 1x 4.2 17.4 41.5 model | log
R-50-FPN Faster pytorch - dpool 1x 5.0 17.2 38.9 model | log
R-50-FPN Faster pytorch - mdpool 1x 5.8 16.6 38.7 model | log
R-101-FPN Faster pytorch dconv(c3-c5) - 1x 6.0 12.5 42.7 model | log
X-101-32x4d-FPN Faster pytorch dconv(c3-c5) - 1x 7.3 10.0 44.5 model | log
R-50-FPN Mask pytorch dconv(c3-c5) - 1x 4.5 15.4 41.8 37.4 model | log
R-50-FPN Mask pytorch mdconv(c3-c5) - 1x 4.5 15.1 41.5 37.1 model | log
R-101-FPN Mask pytorch dconv(c3-c5) - 1x 6.5 11.7 43.5 38.9 model | log
R-50-FPN Cascade pytorch dconv(c3-c5) - 1x 4.5 14.6 43.8 model | log
R-101-FPN Cascade pytorch dconv(c3-c5) - 1x 6.4 11.0 45.0 model | log
R-50-FPN Cascade Mask pytorch dconv(c3-c5) - 1x 6.0 10.0 44.4 38.6 model | log
R-101-FPN Cascade Mask pytorch dconv(c3-c5) - 1x 8.0 8.6 45.8 39.7 model | log
X-101-32x4d-FPN Cascade Mask pytorch dconv(c3-c5) - 1x 9.2 47.3 41.1 model | log

Notes:

  • dconv and mdconv denote (modulated) deformable convolution, c3-c5 means adding dconv in resnet stage 3 to 5. dpool and mdpool denote (modulated) deformable roi pooling.
  • The dcn ops are modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch, which should be more memory efficient and slightly faster.
  • (*) For R-50-FPN (dg=4), dg is short for deformable_group. This model is trained and tested on Amazon EC2 p3dn.24xlarge instance.
  • Memory, Train/Inf time is outdated.