Skip to content

Official pytorch implementation of paper "Inception Convolution with Efficient Dilation Search" (CVPR 2021 Oral).

Notifications You must be signed in to change notification settings

yifan123/IC-Conv

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IC-Conv

This repository is an official implementation of the paper Inception Convolution with Efficient Dilation Search.

Performance

Human pose estimation results on COCO

Getting Started

Download ImageNet pre-trained checkpoints from Google Drive or Baidu Drive with password gq87.

Extract the file to get the following directory tree

|-- README.md
|-- ckpt
|   |-- detection
|   |-- human_pose
|   |-- segmentation
|-- config
|-- model
|-- pattern_zoo

Easy Use

Users can quickly use IC-Conv in the following simple ways.

from model.ic_resnet import ic_resnet50
import torch

pattern_path = 'pattern_zoo/detection/ic_resnet50_k9.json'
load_path = 'ckpt/detection/r50_imagenet_retrain/ckpt.pth.tar'

net = ic_resnet50(pattern_path=pattern_path)
state = torch.load(load_path, 'cpu')
net.load_state_dict(state, strict=False)
state_keys = set(state.keys())
model_keys = set(net.state_dict().keys())
missing_keys = model_keys - state_keys
print(missing_keys)
inputs = torch.rand(1, 3, 224, 224)
outputs = net.forward(inputs)
print(outputs.shape)

For 2d Human Pose Estimation using MMPose

MMPose users can use IC-Conv in the following ways.

  1. Copying the config files to the config path of mmpose, such as
cp human_pose/config/ic_res50_k13_coco_640x640.py your_mmpose_path/mmpose/configs/bottom_up/resnet/coco/ic_res50_k13_coco_640x640.py
  1. Copying the inception conv files to the model path of mmpose,
cp human_pose/model/ic_conv2d.py your_mmpose_path/mmpose/mmpose/models/backbones/ic_conv2d.py
cp human_pose/model/ic_resnet.py your_mmpose_path/mmpose/mmpose/models/backbones/ic_resnet.py
  1. Running it directly like this.

Model Zoo

We provided the pre-trained weights of IC-ResNet-50, IC-ResNet-101and IC-ResNeXt-101 (32x4d) on ImageNet and the weights trained on specific tasks.

For users with limited computing power, you can directly reuse our provided IC-Conv and ImageNet pre-training weights for detection, segmentation, and 2d human pose estimation tasks on other datasets.

Attentions: The links in the tables below are relative paths. Therefore, you should clone the repository and download checkpoints.

Object Detection

Detector Backbone Lr AP dilation_pattern checkpoint
Faster-RCNN-FPN IC-R50 1x 38.9 pattern ckpt/imagenet_retrain_ckpt
Faster-RCNN-FPN IC-R101 1x 41.9 pattern ckpt/imagenet_retrain_ckpt
Cascade-RCNN-FPN IC-R50 1x 42.4 pattern ckpt/imagenet_retrain_ckpt
Cascade-RCNN-FPN IC-R101 1x 45.0 pattern ckpt/imagenet_retrain_ckpt

Instance Segmentation

Detector Backbone Lr box AP mask AP dilation_pattern checkpoint
Mask-RCNN-FPN IC-R50 1x 40.0 35.9 pattern ckpt/imagenet_retrain_ckpt
Mask-RCNN-FPN IC-R101 1x 42.6 37.9 pattern ckpt/imagenet_retrain_ckpt
Cascade-RCNN-FPN IC-R50 1x 43.4 36.8 pattern ckpt/imagenet_retrain_ckpt
Cascade-RCNN-FPN IC-R101 1x 45.7 38.7 pattern ckpt/imagenet_retrain_ckpt

2d Human Pose Estimation

We adjust the learning rate of resnet backbone in MMPose and get better baseline results. Please see the specific config files in config/human_pose/.

Results on COCO val2017 without multi-scale test
Backbone Input Size AP dilation_pattern checkpoint
R50(mmpose) 640x640 47.9 ~ ~
R50 640x640 51.0 ~ ~
IC-R50 640x640 62.2 pattern ckpt/imagenet_retrain_ckpt
R101 640x640 55.5 ~ ~
IC-R101 640x640 63.3 pattern ckpt/imagenet_retrain_ckpt
Results on COCO val2017 with multi-scale test. 3 default scales ([2, 1, 0.5]) are used
Backbone Input Size AP
R50(mmpose) 640x640 52.5
R50 640x640 55.8
IC-R50 640x640 65.8
R101 640x640 60.2
IC-R101 640x640 68.5

Acknowledgement

The human pose estimation experiments are built upon MMPose.

Citation

If our paper helps your research, please cite it in your publications:

@article{liu2020inception,
 title={Inception Convolution with Efficient Dilation Search},
 author={Liu, Jie and Li, Chuming and Liang, Feng and Lin, Chen and Sun, Ming and Yan, Junjie and Ouyang, Wanli and Xu, Dong},
 journal={arXiv preprint arXiv:2012.13587},
 year={2020}
}

About

Official pytorch implementation of paper "Inception Convolution with Efficient Dilation Search" (CVPR 2021 Oral).

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages